MidJourney MASTERY Beginner's Guide - EVERYTHING You Need to Know! (AI Art Tutorial 2023)

mid-journey beginner's guide for 2023 we are entering Quarter Two of 2023 here actually I guess we're fully in it it's been a while I've been teaching mid Journey for a while I've been using it for a long time I use it on a near daily basis this video is going to contain everything you need to go from discovering what it is to using it professionally on some crazy level and unlike a lot of the videos that say from zero to Pro or something like that there's a lot of Secrets baked in this video instead of doing a big multi-part course like I did for my original beginners guide which is still super useful if you want to watch it Link in the description I'm just gonna pack everything into one video so we got time codes in the description chapter markers on the YouTube player and I'm even gonna have some on screen this is going to be intense this is going to be awesome and I have so much to share with you let's jump in and first answering what mid-journey is if you somehow are discovering this video and and don't know no shame in that but it is at this point a paid product which is unfortunate mid-journey is an AI art generator a tool that uses a process called uh clip which is sorting images based on what they are the Styles within them the type of medium that they were made with and diffusion which takes a gaussian noise pattern just random noise and tries to create those eye those objects those subjects those styles from that noise and then it trains itself to do that unlike other models like stable diffusion and things like that mid-journey is constantly being retrained on itself and they have a lot of wizard rehabiting behind the scenes with regards to how they manipulate the prompt you give it to give you the best results which makes it incredibly powerful and a lot easier to get up and running with you know beautiful basic results but it there's a lot you can do to really manipulate and control it and that's what we're going to get into here to use mid-journey you will need a Discord account Discord is a video calling voice chatting text chatting client kind of like the old Skype days uh TeamSpeak IRC channels if you've ever used those like a big instant messenger client and there's Discord as a platform as its own thing and then you have individual servers based on interests companies Brands I have my own server for my community things like that and you will need a free Discord account to join with mid-journey so you click sign in here it will ask you to sign in with your Discord account or make one if you haven't already I'm going to go ahead and sign in here I love the easy QR code scan login it's going to ask you for permissions to recognize your account give you access to your server it specifically says what it can and cannot do sorry if you wanted it to make a mixtape and then it's going to throw you at the mid Journey website and you're going to have nothing here if you haven't done anything with it and you're gonna be like what the heck what am I supposed to do here mid Journey actually operates through this Discord client instead of having a web interface which they are supposed to be working on it operates the bot interacts with Discord itself to make it easier to use but also I guess to give them less Dev work on the website again it's coming soon and so if you are just on their website none of this is helpful to you right now so what you need to do is go to your username at the bottom here which should be your Discord username click the three dots and choose go to Discord this is going to take you to their server I use the desktop client for Discord instead of the web browser version you don't have to download anything you don't have to use the you know the the desktop client if you don't want I do because it's easier for me to show you things and because I use it on a daily basis also if you get taken here to this old FAQ Channel and it's like whoa messages failed to load it's fine the their invite is just set up to take you to a channel that doesn't exist anymore you got all the stuff on the side I'm going to switch to my full Discord view here this is what Discord looks like when you're really using it got friends lists server lists are these these circles over on the left so I have a art category with mid Journey here and this is the mid Journey server so the far left are the servers that you're in so you might be in some others we're going to make a private server in this video so that you can organize your stuff the to the right of that is going to be the channel names of the different channels available in the server so the analogy I used in my last course was that Discord is like an office Park the individual Discord servers are like buildings in that office Park so you might have a building that's primarily occupied by a bank or by support for a restaurant or something then the channels are going to be the individual floors or Departments of that building with there's threads and other kind of micro conversations that would be the individual rooms so we're going to go ahead and just go to the basic announcements thread which shows all of the updates for the program so for today they introduce new billing email address you can reach out for billing issues they have a new style for their anime version of mid-journey that we'll cover as well and there's some new moderation tools coming that is actually pretty exciting so you've got lots of information here you got recent changes to Mid Journey on the whole which gives you updates about new features separate from the server announcements you got rules about being on the server basic stuff they will often pop up polls with regards to getting feedback for how to direct mid-journey you can join in there you got rooms for support and then you have the rooms that you can jump into to start making prompts and we're going to jump in here it's not going to let me use the newcomer rooms because I am not a newcomer but we're going to jump in here and show you what that experience would be like so you scroll all the way down this is going to be super overwhelming at first but do not get stressed we're basically going to leave this area and do it in your own private sandbox instead of dealing with all of this in a moment but I want to show you what kind of the expected first steps are so you go to one of the Newbie rooms you'd have to subscribe to a mid Journey now because the trial is temporarily disabled in order to use it so you'd have to go to your subscription settings and sign up for a plan or something like that they do have a cheap 10 a month one which would be the way to go for most of you is just starting out I'm on the big Pro Plan because I do a ton of generations and use my photos for stuff and all of that jazz but at this point you would have to be on the paid plan to use it Unfortunately they have disabled the trial but you hop in here to the Newbie room and you just start typing so I'm going to walk you through we're just going to generate our first image so you type slash which gives you a bunch of different commands for different Discord Bots Bots are just like services that run through Discord that you can interact with to do certain things and so it gives you a whole list of a bunch of them for different things that might be in the server that you're in but we're just going to type slash imagine all of the main image generation in mid journey is created through this slash imagine command we're gonna click prompt so we're just gonna throw something you know at the wall and see what sticks so I'm gonna say a very happy puppy sitting by a park bench the puppy is a very good boy smiling waiting for pets and we sent now it gives you this message here showing that it received a prompt it's waiting to start and it's going to generate the image that we have requested now what you will quickly discover in these public rooms is that it's going to start showing you this and then a bunch of other people are probably going to start generating images and yours is going to fly away yep it's gone it it's all gone my image cannot be found it can it can there is a button in the top right of the Discord user interface here this little inbox button looks like an old paper tray inbox oh but ours just popped up here too oh look at the good puppy little good though good puppies they're so good if you don't see it or you get a little red indicator over here saying your message was made and you can't find it click the little inbox icon and you're gonna have unreds mentions things like that it's going to have your most recent thing is going to be the most recent person who tagged you which if you're just using the the Discord bot here it's going to have your message and then you can jump to it and we can start interacting with it so you have some controls so by default mid-journey generates a grid of four images for you to work with and these are going to be lower resolution but you've got four images that you can consider that you can use that you can whatever now you can save this out on its own like you could just save right click save image copy it into your photo editor whatever these are going to be low res they're not going to be super useful but if you just want to grab those that's fine or you have the option to upscale any of the individual images so you got U1 two three and four and that's in the grid of one two three and four and so you can upscale any one of them so I'm going to say upscale too but then you also have V1 V2 V3 and V4 these generate another grid of four Images based on this one as an idea to generate more like that so let's say I upscale number two because it's just perfect and I want more light or you know I want the high res copy upscaling just means to make bigger it's going to render it with more detail at a higher resolution with the newer V5 model it doesn't change anything about it it just spits out the direct copy because it renders that way by default we'll talk about that more in a little bit but say version one here this this first one I really like but I want to see if it can do something else with it I can click V1 Now by default it's just gonna fire off for you for me it's going to prop ask me for any additional prompt stuff because I have remix enabled which is a really cool feature we'll get to it we sent off those commands but again the server is packed I have no clue where anything went but I have these pings here in my mentions where you can watch the progress go which is always really cool you can watch the images as they generate here basically watching the computer think assuming you can keep Discord in the same spot long enough you can't always it's also just because it just finished it and here is our variation we have four more images based on that one cute image that we can continue iterating on or upscale one of these you also have a little swirly button here which just will send off that same command to generate a new set of images based on the same idea in case you want to try again in case it didn't quite land you know that kind of thing we also have our upscale that we were waiting on we can jump to that message and now we have a big high resolution copy of this image that we can again save out we can view on the web on the mid-journey website which we will come back to in just a second you also have some options for different versions of upscaling you can make more variations like the V1 through V4 buttons here so if I really really like this but I want to see what more look like and click make variations it would generate four more I can also Mark it as favorite which will show up on the website for you to keep track of some of your favorites so that is the basic process of using mid Journey that is it you just type slash imagine and whatever you want to generate and it will take a little bit and spit it out now we're gonna go take a look at the mid-journey website real quick I want to go ahead and give you a disclaimer that in this course people get really annoyed by this but it's by Design we're going to bounce around a little bit so right now we just generated our first image instead of just overwhelming you with all kinds of additional prompting tips and all of that we're gonna go take a look at the website so you can see what you can do with your images and I'm going to bounce you back and forth what I consider a natural path of discovery of using the journey and getting the most out of it as you go as to not just throw way too much info dump at you up front so you can you can find a natural organic way that you would navigate through using mid-journey here so you can see here we also got whatever you want to generate it tried to put in a bunch of text here which mid Journey does not produce great text but we did get I really like this image I don't know what it is but I like it I kind of like the dude staring at the mirror he's a little a little creepy but I can mask out the text and do something with that the others aren't super useful to me so I am for fun gonna go ahead and upscale one and two and I will save those for later but as mentioned with one of our upscales which I have now of course completely lost there it is we can view it on the website and that takes us to the image page on the website every image you generate with mid-journey is not just in Discord it's stored on mid-journey servers so it's available on their website through your profile you just have to log in with the same Discord account used to generate it and then on your profile page will be virtually every image you have ever generated with mid-journey I generate my images in What's called the stealth mode meaning that it's not displayed publicly on my profile unless I want it to be which means if you click on if we go back to our original link here if you just click on your user profile link here this takes you to your public page or I haven't generated stuff in a very long time however if you click home in the top left it takes you to your private home page which has all of your Generations here which is really cool you can see here those two we just upscaled that good boy puppy we just upscaled and a bunch of stuff for some visualizer projects I was working on so all of your images are going to be here if you ever lose track of them in Discord if you ever delete a image or a Channel or something in Discord you'll have them here now you have a lot on Mid Journey's website and I want I want to go through the basics of using it just because I think the more you use it the more you will get a greater experience out of mid journey in the first place so you can view your images you can sort by the upscales so you're just the final products the grids themselves if you want to go back and like pick one from them or you can view them all you have on the individual images the ability to hide them from View or rather publish them from interview if you are in your private generated stuff that we're going to talk about you can rank rank your images ranking is super important because it's hell it helps mid-journey train its algorithm by ranking images higher it tells the algorithm that this is a favorable result for the prompt that it was given I'm not going to rank this one because it wasn't super really related to the prompt actually I'm not going to rank either of those but I am going to rank this upscale because it was a great one so I'm going to say yes I love it it did a great job a poor ranking with the you know the throw up emoji would be saying hey this didn't do a very good job and it helps over time train the new versions of their image generation generating algorithm as they see fit if you try to abuse it by ranking wrongly back to back to back or whatever they they will ban you for it so you have that ability you can click on individual images take a look at them you can see when it was generated uh some of the flags used to generate some of the commands and things like that you can see related images generated by other people which is really handy for improving your prompts as well as just finding inspiration you can save the image I'm just going to save the PNG to your local storage takes a second to fire off for some reason I'm not a huge fan of that but there we go it downloaded you can add it to your favorites which just helps you find it for later it doesn't change anything about how anyone else sees it you also have these three dots which allow you to open the image in a new tab or find it on Discord which is really handy you can use it as your profile picture or your cover Banner you can report ones that are either like harmful or contain nudity or something like that that are against mid Journey's rules you can create collections to group different images together for you know collections like you would with normal art or just organizing things by folders you can also come in here and copy a lot of information about this prompt you can copy the full command which is going to give you exactly what you use to generate it so if we come in here we do slash imagine I hit tab whenever that pulls up so I don't have to click Ctrl V to paste gives me exactly everything I typed in to generate this prompt which includes a little up beta for the beta upscaler which we'll talk about in a bit you can copy just the prompt which is not going to contain that little parameter at the end so I can say slash imagine paste that in we just have the text you can also copy the seed and the job ID which will be used to identify the image in the information used to generate that image later on so lots of powerful stuff you can do there just on your own images and you can go in here and click this rate jobs button and it will take you through all of your past generated images on a queue to rank them to help again improve the algorithm and train things you can also go over here to explore and explore lets you see the entire communities set of generations which allows you to get tons of inspiration see what other cool stuff people are doing rank them to again help train the algorithm keep track of what you want for example these blue flowers are absolutely gorgeous I'm going to add it to my favorites and I'm going to rank it very positively now let's see what else is this these are just the hot ones so these are going to be like the best of the best I love this Noir style I'm gonna I'm gonna rake it very high and I'm also going to add it to my favorites and once you have favorited some images you can now view them over here in the top column where it says favorites and this way you can come back to them to analyze their prompts to take ideas from them and so on again the the image I just favorited did not show up mid Journey's website is not the most stable thing it should be here it's not at the moment sorry I'm not responsible for it and they even have a favorited category under favorites it's something but mostly you stick to the community tab you find ones you like I love this cat painting this is the coolest thing ever I'm gonna rate it highly I'm Gonna Save it but also if you want to start working with the prompts that other people have used you can click the three dots and copy the you know full command or prompt like I talked about and instead you can just do slash imagine paste that in there you can see everything they worked with they put in the aspect ratio twice which we'll talk about what that means in a little bit so I'm going to delete one of those and we can fire that off and see if we get a similar looking image to what they did or we can start iterating on it say we like this but we want a slightly different style or we want to change something about it we can do all of that just by looking at the images and you know manipulating them and then you can still see similar you know related Ones based on the prompts since it has all of that data it can find you similar looking images and work with that it is really powerful to go through here both for inspiration for getting prompt tips to see what how people achieve certain Styles or you know produce certain characters and by going through and ranking images you help improve the image generation process in general so I highly recommend when you're not you know when your Sparks for ideas you're not generating something yourself spend time here there's also a rank pairs thing to where you can spend more time in a queue ranking stuff to you know help improve the algorithms and it tries to build like similar looking images so you can rank which ones are better so it knows how to like favorite things this is a really cool aspect of it like I said it's one of the big advantages of mid-journey now you can follow individual creators if you like so we can scroll down find something we think is really cool here we go cliche like vaporwave thing maybe I'll maybe I like that this person made this I'm going to look at their profile this person has 205 000 generated jobs with mid Journey they have spent a long time in there so much so that the website's having trouble showing it you can see their collections here at the top let's say I really like their work the stuff they've been doing the Styles they achieve I can click follow and now when I go home or rather now when I go to explore I can click the follows Tab and it's going to show me the most recent images in a feed just like the normal home page but from the people I follow I really hope they get their web service together there we go now you can see a bunch of stuff from people that I follow which is a better way to curate this kind of Explorer feed now we can go back it has finished generating our image but again we lost it so we're going to click the inbox icon we're going to find it now this was using that cat prompt that we found and it has generated four images that are kind of similar based on what you know the prompt comes up with when you feed mid Journey a prompt you're not going to get the same result for the same prompt every time that's not how AI art generation works and so it's a good thing because that means you're never going to make like a carbon copy of someone else's unless you go out of your way to do so but it means like if you want a specific style you will have to work with it I really like how it produced number two here and kind of number one someone else scale both of those and we're gonna move on now remember if you fire off an upscale job or a render job or something you don't need to act on it immediately it will stay here in Discord and you can find it in your mentions and at any point in time you can again go back to your home page on the mid Journey website refresh it and eventually find what you generated doesn't update immediately it used to it used to so next let's talk about the anatomy of a prompt what makes up an image prompt what are what are the parts of it it's mainly divided into kind of three things excluding the Imagine command so you have the slash imagine that you type out click here tap here it works on phones too there's a Discord app for phones then you have your description your your whatever you want it to be for example this person's prompt was beautiful woman cyberpunk style pink and purple that that would be what it is so we're going to say cyberpunk woman and leather jacket and rainy Tokyo neon Alleyway mid Journey does really well with the cyberpunk Styles that's why everyone does it all the time I know it's really overplayed at this point but it's popular aesthetic I'm a huge fan myself we got a smiling 51 year old man playing a grand piano piano in a beautiful Meadow that's someone else's that is the descriptor of you know the subject of what you want and typically that's broken up into a more like direct subject oriented description in plain text and then there's going to be some commas and then just a bunch of extra descriptive material so right now I just have cyberpunk woman in a rainy Tokyo neon Alleyway but then you see here they have a bunch of stuff with commas and so I could say long pink hair wearing headphones with blue eyeshadow red lipstick that kind of thing and then the third part of your prompt is the parameters and there are parameters for many many different things and we're going to cover a bunch of them in this you know guide but one of them is for something like the aspect ratio to generate a different width or height of your image instead of the default one to one square size you've got seeds you can use you got a chaos element that can introduce some Randomness to your prompt to make it more interesting you've got a quality slider which takes a lot longer to generate and uses up more of your cycles for your subscription but produces a better image lots of little things like that and so for this I'm going to say dash dash AR for aspect ratio Space Two by three because two by three is more of a typical uh portrait it's if you have you know a standard film camera it's a three by two sensor or film slide I guess large form or medium format maybe not but then if you turn it vertically like you would for a portrait becomes two by three I have some various aspect ratios showing on screen right now obviously a standard computer screen would be 16x9 this video you're watching now is in a two by one aspect ratio your smartphone would be 9x16 typically those are aspect ratios so we're going to fire off this command and ultimately that is what a prompt is consisting of that your imagine command your dis your subject description your your plain text for what you want maybe some additional qualifiers or descriptors and then your parameters for manipulating the image generation process that that's the anatomy of a prompt that is what you need to know to honestly play a lot and have fun with them in the first place I also wanted to highlight here while we are waiting on our image to generate you can interact with other people's images here so and they might think you if you're upscaling one that they didn't want to spin it upscale on but like for example maybe I like this realistic photography of you know Demon Slayer people maybe I want number or the hands are weird on these so I guess number two I can go ahead and upscale this or I can go ahead and make a very variation of number two and make more like it and continue iterating however much I want it's a very open and like interactive creative process which is really exciting all right here we have our our rendering of a cyberpunk woman in a leather jacket in Rainier neon Tokyo Alleyway with long pink hair wearing headphones with blue eyeshadow and red lipstick we have our two by three aspect ratio for each of the four images and for the most part it kind of stuck to our descriptor you can't really see the blue eyeshadow on most these the blue kind of went into the hair but for the most part this is what I want like this is a pretty cool style I'm going to say we upscale one and three and maybe four only upscale the ones that you really want to keep otherwise save them for later and decide later as you go through your iterative process it is a lot easier for me to just go ahead and upscale them and work with images individually to figure out where I want my creative process to go but not everyone's going to be that way there's the variations we generated on that Demon Slayer you know image you can see here our upscales waiting to start and it will tag us when they're ready so that was the basic anatomy of a prompt but there is as I showed a lot of different things that you can add to Any Given prompt to help better describe what you want to create different styles you can you can really build anything with the mid Journey prompts and feed it so much information about what you want and we didn't even you can actually use already generated pictures as references to help better direct the art that you want generated to here's one of our upscales yep they are coming through now this is what the older mid-journey V4 model there's a new image model that they generate Things based off of called V5 that produces much more photorealistic images by default V4 is still what it's set to because V5 is still in testing and so these don't look the most photorealistic they look a little bit more hand-drawn color pencily because that's the default kind of mid Journey style but the new model looks a lot more photorealistic by default next up you have the the extra things I mentioned that you can inject into your prompts to produce better images these are Styles flavor text Art mediums um the purpose of the thing you're trying to create you can do uh composition camera choices all sorts of things like that and before we get there I'm going to show you my first little tidbit we're going to dive in a lot into like working with my journey more independently and away from all the noise in a little bit but I'm gonna show you my first little tidbit for working more privately in that you can direct message the bot so the mid Journey bot that delivers your images you can click on its name and you can add it to your own server which we'll talk about in a minute or you can just message it I'm going to send hi so it takes me to my direct messages UI for Discord here it's not going to respond when I say hi but now we can generate our images here which is a little bit quieter a little bit more focused a little bit more you know oriented around what we're trying to do and so now we can focus a little bit so I will say right now from now on moving forward all of the images that I'm going to generate of the same subject are you going to use the same what's called a seed the seed is just the noise pattern that you that is used to generate an image and so you can tell that a seed to continue to use the same seed over and over to try to get more consistent results since it bases it on the same starting place you don't need to know a town about this just yet but I'm letting you know that I'm going to show you differences by using the same starting point each time let's do something completely different from the cyberpunk aesthetic let's go back to the mid Journey website we're going to explore we're going to look at you know Community favorite images what is something cool we could make I am absolutely in love with this Skyline warping around this bridge going through it with space in the background I'm going to save that image myself and we're going to start with this so let's first analyze the prompt that they use so the core part of the prompt that we've already talked about space science images science images photo library that is the the main thing it's trying to tell mid journey to create which frankly isn't a whole lot but then you have all these other things in the style of surreal City scenes so we told it a style to emulate which is actually a couple style City scenes surreal style then we have some more stylistic you know influence realistic anamorphic art swirling colors as an element to consider trans compilation CD so compilation CD so you can tell it things like for an album cover for a comic book cover for a movie poster you can tell it these things that are the purpose of what you're making as an influence for the design elements it will pick up based on what it's been trained on it hasn't been trained on everything but you'd be surprised that the weird random Grandpa core Forest Punk kind of descriptors that I've never even heard of before in my life and wouldn't believe real until working with mid-journey we got virtual reality we've got a person's name I'm gonna briefly touch on this Joel Robinson I'm assuming that's an artist let's look them up Joel Robinson character featured in the American Science oh from Mystery Science Theater 3000 maybe it's the photographer ah so he's got some very surreal art styles so you can reference other artists that mid Journey might have been trained on to take cues from their style this is one of the biggest controversies of AI art within you know a bro the broader art communities because it's seen as theft or as you know plagiarism whatever I think there is a safe way to use those cues as very minor influences into the style that you're trying to make without really encroaching on their territory or trying to copy them there's the whole quote great art or good artist copy grade artist deal in terms of like you take what you like from everyone that you watch and consume and appreciate in terms of their work and you take little nuggets from them and you combine them into your own unique style it's not cool to just slap you know slap in some artist names and completely rip off their style it's it's not okay and frankly I would recommend whenever possible not using artist names in your prompts at all just to avoid that altogether but that's not going to make everyone happy and it it may be easier to reference people you know their style to build on that to get you a starting point as you're figuring this out for fun anyway just be careful then we have one last little descriptor phrase multi-layered Dimensions which is definitely appropriate here so we're going to click the three dots copy full command we're gonna message the bot we're gonna say slash imagine which we can use here just direct messaging the bot and paste that in so we went over the prompt itself but then we have the parameters like aspect ratio 51 by 91. I get the feeling they fed it like describe I I get the feeling they just pulled this from another image we have a quality of two a stylized parameter which we haven't talked about yet of 750 it goes from zero to a thousand and then the new version five for the sake of clarity of what we're talking about right now we're just going to focus on the prompt itself and then I'm going to say seed of eight six seven five three oh nine got it I got it all right so this is the prompt that the person used we're just going to generate a basic one by one image we're gonna fire it off but we're also going to fire one off where we add more to it so I'm going to hit the up Arrow which is going to recall my most recent prompt if that doesn't show up Ctrl Z often pull offs also pulls it up and then you click the prompt button there so now we have our prompt with all of these different style influences but maybe we don't want it to look like Joel Robinson's art maybe we don't want it to look like a trance compilation CD maybe we don't want swirling colors I kind of like the rest of it but maybe we want four Doctor Strange because it had the like City folding on itself kind of dealio going on maybe we want a cinematic sci-fi film still frame so that is one of the things you could use is we want to still frame from a cinematic film of some sort maybe we want it to be more to to with alien spaceships in the sky so we can add in these extra little elements of what we want what we want to show up and we can add an extra stylistic you know influences so we're gonna show the difference between what it just generated and again we're using the same seed and adding in these extra descriptors in place of the ones from the original prompt and you can see here we're already getting something different from the original posts we were looking at both because we are removing some of those extra chaos elements and stylized elements and stuff that we haven't covered yet but also because we're using a different aspect ratio the aspect ratio again that's the dimensions of with the height and I have a video on my main Channel about this if you need it linked below that actually has a pretty big influence on the composition of the scene that you want so for example this has a long you know winding little Trail here here as well as really tall buildings and a sky it has this really tall composition because it's a really tall aspect ratio the default one-to-one aspect ratio is very square and so it has to change you know where the elements might go we'll be covering aspect ratio more in a little bit all right now we've gotten something different by swapping out some of the descriptors of the artist the trance compilation CD swirling colors and things like that but specifying instead that it's for Doctor Strange that we want a cinematic still from a you know from a film that we want alien spaceships in the sky we get those alien spaceships we get something a lot more like epic in terms of scope of the scene to be more cinematic and we've got something a lot more sci-fi versus just kind of City bint but now we can take it a step further and manipulate it even more so we're still using the same seed I'm going to click tap my up Arrow to bring that prompt back actually that's going to take me now that I have already sent that prompt it's going to take me to edit my last sent message which is hi we don't want to do that instead I'm going to do Ctrl Z which is going to give us that prompt so now we want to manipulate it even more so Maybe instead of just a sci-fi film still frame we want for 1981 movie poster so now we get something that looks a lot more like what might be expected on an old movie poster we've got that hero character in the middle we've got a lot more of that like framing elements going on here we have a little bit of text because movie poster and album covers often prompt text out of mid Journey frustrating you can do dash dash no text at the end to try to combat that doesn't always work we have something a lot more kind of moving towards the center line of focus for a movie poster and it usually starts adding in like Hero characters and things like that so those are the different like style influences that we can have well there's a million of them I have other videos on my channel if you want a list of style influences and actually I'm gonna have some resources we'll talk about in a little bit but those are the ways that those uh you know influences could impact your image as one more example here just to show you this kind of thing we are going to pull back up that prompt and instead of four or instead of cinematic still frame 1981 movie poster we're gonna say Gothic painting in Cathedral everything else is the same it's still even for Doctor Strange it's still all the Sci-Fi elements still got alien spaceships in the sky but instead we're going for a Gothic Cathedral painting style now this didn't this didn't actually change the medium of the image in the way that I intended we would have to kind of completely reword The Prompt but you did see here that the cathedral and Gothic style did heavily influence the subject of our creation we still have that warped perspective and the grand scale with the planets and the spaceships but now we got Gothic Cathedral Castle style architecture instead of the the New York City landscape or the Sci-Fi buildings this is what I'm talking about you can combine so many things and so next I want to show you the the means of doing style Fusion for your images so we're going to scroll back up go with our original prompt here seed and everything I'm going to do slash imagine paste it in now we have the in the style of surreal City scenes so I am going to delete this and we're going to say style Fusion of surreal City scenes and Renaissance I spelled that wrong whatever painting did I spell it wrong I don't think it did and Renaissance Gothic painting so we're using style Fusion surreal City scenes and Renaissance Gothic painting same seed same everything else as our original prompt here which gave us our very city-esque surreal look but this time we're fusing those two Styles you can do this with all manner of styles if you do this or what's called multi-prompting which we'll get to you can combine multiple different styles together you can do this this with different artists different artists that you're taking inspiration from to create something wholly unique by referencing you know one style or one artist style and another artist style that look completely different and combining them together with your own ideas of extra stuff to make it your own you can really bash a lot of things together with mid-journey now you see here we still have the surreal style we still have this crazy warped perspective uh this third one is very very similar to the original image we were taking inspiration from but most of them are taking on a little bit more of a painterly style we've got those Renaissance little clouds around this one the color scheme has changed we end up with a a subject person again for some reason but we're definitely taking a little bit of an influence there we can do something really crazy copying all of this again or just hitting Ctrl Z space science images photo library we're going to say style Fusion something I did forget to do was take out the Joel Robinson name as well we're gonna say style Fusion surrealism mixed with hyper realistic surrealistic surrealist painting mixed with Hyper realistic 3D render something you want to note when iterating is you want to change as few variables as possible each time you go that way you know you're going in the right direction and that you know the change that you made actually had an influence on what you're trying to do this is a bad habit of mine in that I will change everything all at once because I have ADHD and struggle to you know do things slowly one step at a time I just want to reach in and get the exact changes that I want but if you continue to tweak one little element at a time especially when using the same seed which again we'll talk about a moment don't worry you're gonna much better navigate a path towards the image that you're actually after now we've still got the swirling colors we've got the surrealism but now we're just completely off into nowhere I did change a lot all at once but you can see my point here is ultimately to just show you the influences of the different styles on what you can do and again there's no end of list of them I'll flash some up on screen with little comparisons super quick we'll cover more and I'll even show you some resources in a little bit time coded sections in the description for where you can get lots of more tips for this but honestly just spend a lot of time go I mean we found this image did I save it I hope I saved it yeah we did spend a lot of time going through the mid Journey Community Feed and finding images that just stand out to you that you want to take inspiration from because there are so many and it will really help you navigate like what exactly you're after and you can take prompt tips from them by seeing what they have I like a lot of these I'm gonna save them just for inspiration for later although I was more going for the city skyline kind of look in the first place not sci-fi that shot's really trippy all right we have gone through we've made some images we have learned about prompting and style influence and how to recall our last prompts really quickly but we have a bunch of images to work with and we are either just dming the bot with like an endless feed which is really hard to keep up with or perhaps you're still in the various public channels on the mid Journey server and you know there there's it's so hard to keep track of what you're doing and stay on task this way so let's talk about some organization tips for your files and for mid-journey to get the best results first and foremost I recommend having your own private server for Mid Journey so that you can organize things going back to that office building analogy I mentioned before where your servers are your individual Office Buildings and the floors are your different channels here if you're just taking up space in someone else's office it's really hard to get you know to spread out to have your stuff protected to to have your creative space and have your thoughts you know aligned and focused and you want your own space for that and thankfully making a Discord server is free so in your server lists on the left you probably don't have as many as I do scroll all the way to the bottom and click add a server I'm going to say create my own for me and my friends default name whatever's fine especially if you don't plan on sharing it or inviting anyone you don't have to change anything you don't gotta personalize it you just click create and then you want to make a bunch of channels for different things I'm not going to keep using this one because I've already made one and I want to walk you through what you should have but theoretically you should have something a little something like this now I have a lot don't let it overwhelm you but these are different so we have categories categories would be like the floors whereas the channels would be the the offices so I have categories so for example I have merch ideas granted I have a scrap channel in there maybe I should not like make it its own thing there we go but I have merch ideas so specific Generations that I'm going to work on or inspiration for creating merch so stuff I want to keep track of there I've got stuff for client work so again it could be inspiration or it could be Generations that I'm doing for example for YouTube thumbnails for album covers and visualizations I'm doing for our Music projects at backingtrack.gg I've got a category for my personal projects so I've got stuff relating to like storytelling I've got stuff I was trying to generate for RPG Maker for game assets I've got stuff for Dungeons and Dragons I want to work at got stuff for glitch Arts thing I wanted to generate images for me to run through all my analog glitch gear to make glitch art on a CRT with and then I have in it a massive what I call inspell category which is just all the things that I find around the internet for inspiration think of it like a Pinterest board you could just use Pinterest and in fact I do use print Pinterest pretty like well off and on religiously when I'm when I'm actually into it I use it pretty heavily so my inspo channels on Discord would be very similar to my different Pinterest boards that I would have for different things and so I've got one for design ideas I've got one for portraits for glitch art that I find you know on Instagram on Reddit on whatever I'll either take screenshots and directly upload them or I'll just link the post if I'm in a hurry all sorts of inspiration that I either want to learn how to do take inspiration from take little bits and pieces from things that just inspire me either from my studio itself or for my style to make you know what I want to make and organizing all your stuff in your different channels and it's super easy to make a channel you just right click create channel this is a channel you can do dashes or underscores for spaces great Channel at all you want in terms of non-mid Journey related stuff for your inspiration and stuff that way you can have it all in one place because you can use reference images as I alluded to before and so you want a place to kind of upload these easily to click on them and copy the links to them this will help you out immensely both for your generations and for just your overall creative space think of this as your as your canvas as your drawing desk whatever you want it all in one place so make your channels for the different things you'd like to do and then we have to invite the bot to our server so go back to your DMs that you had with the bot or find the bot on the mid Journey server delivering images to people click the name and click add to server and here we're going to tell it the Discord server we want to join and click continue I recommend leaving all of these here because they're required to use mid journey and click authorize now I already have it in my server so we don't need to worry about that now when we come over here in our private server slash imagine is right here and ready to go so immediately I can start separating my work into these different categories and I can say slash imagine if you want a t-shirt idea uh horned demon monster profile for business headshot medium shot Gothic horror grainy 35 millimeter film photo on VHS and then since it's a t-shirt we'll do an aspect ratio of two by three say I want to generate that to put it on a t-shirt I can start immediately generating images in my t-shirt category that whenever I come back to my idea for a t-shirt I can either keep working on or copy out of or do whatever and so we can let it generate here for a second now we have some Gothic style profile pictures right here and just right in here I can upscale them I can iterate on that one that one kind of looks like Abraham Lincoln and we can do all our stuff and immediately that's a that's a huge boost to the overwhelming noisy room that everyone else was doing their thing in right yeah we can take it one step further so we have our server we have our category we have our channels and you can also make threads in Discord threads are just like independent conversations that are still on theme for the channel that you're in so for example I'm in a t-shirts Channel but I'm doing a lot of images specifically about this horn monster but maybe I want to be able to generate images about something else and I don't want to have to scroll up forever to find what I was doing so instead of just spamming my chat I can instead click this plus icon create a thread we'll call this demon king shirt we have to put some text in so I usually just copy paste the same idea or I'll copy a link to the or an image of what I've already done hit enter and it's going to pull it up here on the side so if you want to reference other things in your main chat you can sometimes keep that up usually you have to be in the same channel or you can just click on the thread name right here in your Discord settings and you can view it now with how Discord works the threads will automatically archive AKA stop showing up underneath your channel names after a couple days they don't go away though so like for example I have this older thread here for skeleton Kink it no longer shows up underneath the channel name I can no longer immediately jump to it but I can click right here to that thread pull it up I can send a new message or create a new variation of this so I can say make variation I want an aspect ratio of 9 by 16 instead and I want to use version 5. and submit and once messages start being sent in that channel now it pops back up and I can keep working on it you're not losing anything so I highly highly highly recommend setting up your own private server that you can generate all this stuff in to keep it organized get your categories of either inspiration or Generations you want to make threads for specific Generations you're working on and and really sore I do want to have a give a bonus tip here when you're using reference images or variations on an existing image you're beholden to the aspect ratio that it was made before so this is a 4x3 image but it was also inspired by an even like thinner image so we already have black bars on this 4x3 image but it was supposed to be 4x3 and then I told it to make it 16x9 and so all it's doing is stretching up that already weird image you gotta be careful with your aspect ratios we'll talk about those in a moment so the next organization step is of course creating a folder structure of some sort on your computer for when you're saving your images now you can do this any way that works for you I'm going to come over here to my big storage how much actually I'm going to come over here I'm going to make a root folder I'm going to call this AI art for my AR art projects and then you could do a similar layout to your you know channels so for example this could be t-shirts and a folder for patterns and then maybe I want a folder for thumbnails [Music] and maybe I want a folder for album covers and then you can start organizing by year or maybe you want year to be organized in here and for album covers I can do I want just the chill wave ones for the chillwave album then I can come over here and start saving my images so I can come over here to chillwave thumbnail save the images that we were using so I really like this one come over here to web I usually recommend these days clicking the web button and downloading from the mid Journey website because there are some images you will be sent by mid-journey that there is a higher quality version available on the website than what they compress in Discord because Discord compresses images now and so now I can either right click save image as come over here to Windows Explorer click in the field here Ctrl C to copy click in this field Ctrl V to paste takes us right to that folder call this Tokyo it was neon rain save it or I can just click the save button as we did for mini and pull it from my downloads folder obviously if you are in a different browser Firefox by default just dumps everything into one folder other browsers may put them in different places so I can take for example our futuristic sciencey thing bring that back put that in the t-shirts category and we're going to say uh dimensional Skyline as our t-shirt idea paste this as our inspiration image and then I can start manipulating that in Discord to start working with we're going to talk about that in just a second but I want to show you one last organizational thing and that is regardless of whether you're in your private server or on the public server if you ever want an image sent to you just so you have it more clearly or information such as the job ID or seed of a given image you can get that from Discord so for example if we scroll all the way down obviously I showed you how to do these things on the website but if we find something cool I'm a huge fan of abstract art let's see if anyone's made anything cool here these are characters this is not abstract art I another reason to avoid the public server is people rarely stick to a room's given category these are all portraits I realize there's some abstract elements but come on man this person's been going forever pretty sure yeah you can make threads on any public server like my own public Discord server has uh mid-journey added to it on any public server you're doing just make a thread to organize your stuff it's so much easier but just as an example let's say we love these this abstract mirrors image here and we want to be able to work with it we want information about it we want to keep track of it better rather than in the public chat go over here to these reaction buttons that lets you react with an emoji click it and we want the envelope you're not going to see it by default so just start typing envelope and you want the basic envelope click it and when you do you're gonna get a little pink and the mid Journey bot is going to send you that image and it's going to send you the prompt and the job ID now you can jump back to the message that it was from or you can copy all of this including the image to that channel so let's say just for argument's sake I want it in this chillwave thumbnail I can just paste the text in here so I have it and then I want to copy the image with it just so I can keep track of it together here and now I have this message I can reference with all of this information about it and now I can do something really cool I can say slash show and then it's going to want the job ID so we copy that job ID out of here paste it in and that's not going to work if you're not the owner of it but if it's your own images you can I got an email from a viewer who accidentally deleted one of their Discord channels that contained a bunch of their mid-journey Generations their images the stuff they wanted to work with and are freaking out because naturally they lost a bunch of their work good news you don't actually lose anything so I'm going to show you how to recover your work first and foremost there is no way to recover deleted Discord channels so unfortunately that will be gone but you can make a new one you can still access your images though every single image you generate is stored in your mid-journey account so that means you go to Mid Journey's website which is just midjourney.com you choose sign in sign in with the Discord account that you're using to generate your mid Journey images sign in with that account and it will take you to your profile page here it will show you every single image you have ever generated with mid-journey going back to whenever you started with this account and you can sort them by ones that are popular if you've shared them publicly your newest ones you can sort by your oldest one so you can see how bad some of my very first ones were they look like Ms paint drawings it's not the point from here though you can do things with them obviously you can just save out the image itself and photo edit you know do whatever you would normally want to do with an image but with every single image whether you're in the grid view here or selecting a very specific one there is a three dot little menu right here you click that and then you have a couple different options first and foremost if you go to open in and then open in Discord it's going to open a tab with Discord and if that channel still exists it will take you to the message that that image was generated in which is really handy obviously if you deleted the channel it's probably going to error out however you can go over here to copy and you have a few really cool options you can copy just the text prompt you can copy the full command which is going to have all your extra modifiers so first and foremost if we go to copy prompt and then open up here if we just paste this in you can see here it has our image sources and our description but none of the quality the aspect ratio factors any of that if we go to copy command it's going to copy all of those extra arguments and flags you still have to type slash imagine at the front but then you can fire it off and it's going to have your aspect ratios your quality all those other settings however you also have a third thing you can copy and that is the job ID now if you click this click job ID now you have a new command you can use which is slash show and then you paste in that job ID which should be this big string of letters and numbers press enter and it's going to recall this exact generation that was in your Discord before and present it to you boom there we go we have this exact image the exact command the exact generation it doesn't remake it it just calls it back and then from there we can remix it we can remaster it we can upscale it we can do whatever and so you can do the same that with individual images or you can do it with the four race four-way spread so I can go over here copy job ID type slash show paste in the job ID and boom we already have that and then we can still go through like I said variations upscales all of that fairly straightforward Don't Panic obviously if you want everything back exactly like you'd have it you'd have to sort by oldest and copy every single job ID and make a replication of that channel probably not worth it but if you know when you pick out your favorites or there's ones you specifically want to work on you still have them accessible they are not gone so I do recommend spending some time at some point in between generations to filter sort build collections things like that now that it's a feature in my journey if you can but otherwise you have the prompt you have the job ID you could go to the mid Journey website go to explore paste in the job ID hit search now since this is a new one it might not show up but you can see it's right here it pulls up that image and we have all sorts of other jobs relating to it unfortunately it won't bring us that exact image but theoretically once you've found it in here you could then do the same song and dance of copy The Prompt all that [Applause] all right next let's talk about reference images as part of your prompt you can use one or more images as reference for the image that you want to create meaning that it's going to take primarily the colors in the composition of an image that you feed it it can actually analyze it break it down a little bit and apply whatever prompt you give it to those colors and composition to create something new for example I pulled up this image over here on mid-journey of a cyberpunk robot kneeling down in the middle of a post-apocalyptic destroyed City it's holding some colorful balloons maybe I want this to be a clown in a post-apocalyptic City instead of a robot how do we do this well first and foremost we need to make sure we're in the right spot here in mid-journey I'm gonna make a whole new channel now so I can keep track of these a little bit better I'm just gonna right click create channel we're going to call this tutorial that way everything's together from now on so here I'm going to do slash imagine and what you do is you paste the direct URL of the image that you want to use as a reference before you type your prompt so in this case we have one pulled up here on mid-journey we right click and we choose copy image link not copy image copy image link and we paste it in there and then we write out our prompt now a tip if the website you're pulling your image from gives you trouble like it gives you a big gobbledygoop string of letters that doesn't give you a proper file or you have an image locally that you want to use as an image reference you can actually just use mid Journey for this and so I actually recommend this part of the organization figure out what category you want it to go in but create a channel that's just preference images and then you can copy paste the image itself so right click copy image paste it directly into your mid Journey chat and maybe even if you have any sort of descriptor a prompt a I don't know uh just a general idea of what the image or the reason you're using it as a reference image go ahead and paste that in there just so you have it for later this is purely for reference for you mid Journey doesn't see this directly but now I can sit here and click the image right click copy link and now we go back to our slash imagine command paste that in there now we want to get rid of anything after the file extension so dot PNG is the image file extension so I'm just gonna delete that space and then we're going to figure out our prompt so we're going to copy the prompt that this was given now something important to note about reference images that I mentioned earlier in the video you're somewhat beholden to the aspect ratio of your reference image so for example we're going to paste our prompt in here this image was generated with an aspect ratio of 9 by 16. so a vertical you know portrait aspect ratio if we try to generate an image with a different aspect ratio it's going to have a problem so I'm actually uh for reference here I'm going to copy the oh we can't get the seed from it I did not realize that huh I guess we don't need it so I'm gonna do we're gonna ignore all this I am going to use V5 because that's what they use I'm going to use seed we're going to use just a random number that we can keep consistent so we're going to generate a new image based on this one and we're going to go ahead and do what I'm going to show so for the reference image we want to replace this with a clown so we're going to say a cyberpunk I'm going to leave it as cyber no I'm going to say a glitch Punk clown and otherwise we are using the reference image we're using the rest of the prompt the aspect ratio all of that and we're going to generate an image but I'm also going to show you what it looks like when we try to change the aspect ratio so I'm going to say two to one if I want like a cinematic aspect ratio we're using the same seed the same reference image everything and you're gonna see what happens okay I said that and in this case we actually got a wider image I showed earlier in the course when we used the reference image from one of my it's the Skeleton King design here if you if you at least with variations you absolutely end up with some wonky stuff but a lot of the time of course my example that I pull up doesn't show up but a lot of the time you're going to get black bars in your image so it's going to generate the aspect ratio that you want but it's going to put black bars in it in this case it didn't but for example if I you know if I if I make a variation of one of these let's say number one which has the balloons but that's not really a clown there's still no clown here so we're going to say variation one we're gonna say a glitchy clown blah blah blah scary clown we're going to change that aspect ratio to say four by three for some classic retro design now interestingly we have adhered very heavily to the original image prompt here which we can discuss image waiting in a moment so we have not given a whole lot of weight to the clown aspect yet you can see number four here has a little bit of clown action going on but it's still sticking to the cyberpunk robot aesthetic and so we can tweak that with some image waiting which we'll discuss in a moment but you can see here it is very much adhering to our reference image in terms of composition the design language and the colors and so you have to do a lot of work or use what's called image waiting which is basically telling it how much to focus on any particular element of your prompt be it the text or the images themselves to generate what you wanted to see there we go okay so here's our example of our variation we generated changing the aspect ratio so it was originally two by one and then we told it four by three and this one didn't do the black bars that our other examples showed which is very common I see this all the time where you get the black bars and then the stretched and squished aspect ratio and it gets really funky we got really lucky with this one I don't know why but it still is a little squished going from the two to one to the four by three so if you run into that with your reference images or your variations whatever you know you're using as reference causing that effect you can just try to generate a version that has the correct aspect ratio that you want so say for example we wanted that 4x3 aspect ratio we could take our original reference image and just basically build something around it that isn't a solid color to show that that's what we want so in say I'll show you a couple quick super quick ways because a big big message that I'm going to push throughout this little course video here is that the more prep work you do for your images in terms of you know drawing the idea yourself photobashing It Whatever the better result you're going to be able to get so this is part one of this but we're going to dive into it more in a second so if I want a 4x3 image instead of the 16 by or nine by 16 one then I'm going to make a new Photoshop document I'm going to show you how to do it in Photoshop because they have fancy tools to make this better I'm going to show you how to do it in a free app called photop that's on the web uh where you don't need to pay for Photoshop but it won't be as good so if we want 4x3 we're going to say 1440 by 1080 which is a 4x3 aspect ratio we'll talk about aspect ratios in a minute I know we haven't gotten there yet we're building up to it for why you should care all right so I've made a new document that's in a 4x3 aspect ratio resolution doesn't matter a ton you don't want it to be too tiny that details get all crunchy but the journey doesn't really process super high resolution images we're going to paste our image in here get it resized to fit our canvas and Center it then we're going to use the magic wand tool over here on the left select our two blank areas so hold shift to select the second one then we're just going to do edit content aware fill Photoshop has all these AI content aware tools that allow you to manipulate images a lot easier that's part of what you're paying for with those crazy subscription compared to other applications and for this much raw space being generated out of nothing it's not going to look great but it's going to keep it from just having solid bars which will still influence mid-journey's decisions so you can see a little preview over here it's just taking like random details and trying to tile it we can use this brush to kind of get rid of some that we don't want it to include so maybe I don't want this texture off this bag included I don't want the guy included I don't want this thing on the ground included but I'm fine with that background included and I don't want the strings of the balloons so otherwise that should be fine I'm going to go ahead and click OK and it's going to generate a new layer with those sides on it now it's just tiled texture of this stuff again it is not the best what we can do however is perhaps crop a little bit so we're going to keep our ratio we're going to crop in on the balloons just a little bit cropping on the ground a little bit so we have less space taken up by this just raw repeating texture and then I'm maybe gonna make a new layer use our clone stamp tool over here which can clone parts of the image make my cursor bigger hold alt to take a sample and just kind of it's not doing anything at all hold Alt oh because I'm not sampling the correct layer old alt over here and draw in some of that texture back okay make sure at the top sample current and Below layers so you don't have to keep switching layers I don't want the balloons to repeat too much you know fill in some of that building this looks like absolute garbage like I understand that it's not supposed to look like a whole lot it's just supposed to not tell mid-journey to like give you solid bars it's still supposed to look like something of a destroyed City and you could even photobash which we'll talk about later in the course but we can just pull up a city like something like this copy this image paste it in here it's from a video game or something probably and then put it underneath our guy yeah we can use something like that where we just have that City in there as a as a means of giving that information a mid Journey we'll just export this as a quick PNG I'll just make a scrap folder because we don't need this for anything cyberpunk robot destroyed City we'll talk more and more and more about how to get better results from this in the future but that's how you would make that image in Photoshop in a free application if you don't have access to photoshop photop.com this is basically like a web hosted Photoshop that is completely free and so once again we take our we're gonna make a new project make do they have a 4x3 they don't by default so 1440 obviously whatever aspect ratio resolution you were targeting create we're going to copy in our original image controls are going to be a little different since it's Ctrl T is to make a tab we're just going to make it fit here align it and then same thing with that destroyed City image copy image we're just gonna put it behind our guy and file export as PNG so once again we go into our Discord for our reference images we drag in our new crappily photobashed image here upload it now we copy link and now we can use this to generate a 4x3 version so that's how you would handle that use that as your image source instead of the original so to help get some of our clown back we can actually use multiple image sources in our prompt and so I'm going to go ahead and pull up our prompt here and hit Ctrl Z not working anymore all right we're just going to copy paste our original prompt here slash imagine and once you have used an image reference it gives you this s dot MJ dot run URL you can just use that moving forward for your image references you don't have to copy the original link over and over so I'm going to copy the image link for the second clown image and I also pulled up another clown image that we're going to use as well both of them are holding balloons so this is going to help kind of identify that we want a clown here and I'm going to leave it at the same 9 by 16 aspect ratio of the original image so now we have our original original inspiration image which is the the robot holding the balloons that we wanted to replace with a clown we have our prompt where we have modified it to specify a clown instead of a robot and then we have two more reference images thrown in at the front of specifically you know kind of creepy looking clowns holding balloons kind of in the fashion that we want send that off you can see here whenever you use image references now when you fire off the command at first it'll just show those in the preview but they'll go away once it starts rendering the image now you can see now that we have given it sources for what a CL you know for the clown style we're going for we get exactly what we want we get a creepy looking glitchy clown holding balloons in the middle of this destroyed cyberpunk post-apocalyptic City That's The Power of image references reference images however you want to say it it's really cool and really powerful and we're going to keep using it but we need to talk some more about aspect ratios since we covered this whole thing with aspect ratios aspect ratio is the ratio of width to height of your image of video or whatever common aspect ratios include 16x9 for a standard computer monitor or TV 9 by 16 for a standard you know phone screen you've got 21x9 for Ultra wide monitors like I'm using now this video itself is in two to one or eighteen by nine there's 4x3 for old you know CRT video film photography and normal still photos from a you know real mirrorless or DSLR camera are usually taken in three by two or if you flip that vertically two by three I'll have some some shown on screen these you can use to generate you know images for different purposes for mid-journing because if we took the standard Square output of something we have worked with with mid Journey before let me find an example here if we take the standard Square output of say this prompt we just fired off if we just do slash imagine and delete any aspect ratio bits it's going to give us a square image which is a one-to-one aspect ratio this is perfect for posting to Twitter to Instagram social media platforms but if we want to use it for a poster for a video for a thumbnail for wallpapers anything like that it's not going to be super useful because then we're going to have to crop in to get the aspect ratio that we want or stretch it which is going to look horrible and usually you know cropping and things like that won't always be an option because of the composition of the frame so that is one reason you might want to use different aspect ratios is you want to generate something that looks more like you want so for example for like shirt designs I would do something taller because usually they're printed tall in front of just shirt I'll show you some examples of some shirts I've had printed with my glitch art and my AI art that I have in physical form it's pretty cool for certain prints I want something that's like super cinematic and widescreen for like a poster print so I want a wider aspect ratio same thing with posters I had the whole if you haven't seen it I did a whole project of printing mid-journey generated images as posters and like super high fidelity and super big size and it turned out great and you can see some of these here these are all generated in a very tall aspect ratio because that is the size of the poster that I wanted to generate it is 13 by 19 which is the size of the paper that my printer accepts so I messed with that size in this case we do still get basically the same image that we want and so we could crop this if we needed a vertical crop for it but that won't always be the case also the the aspect ratio heavily contributes to the composition cues that mid-journey takes for certain compositions so if we go back to our DMS with the mid-journey bot and take a look at the images we were trying to generate before based on that little swirly Library one if I pull up the original image that we were referencing it is a very tall aspect ratio image it is generated in I think 9 by 16 but it's very tall which affects the composition because then we get these really tall buildings you know warping up into the sky we have this spiraling stared path going through the city we have a lot of compositional Clues because mid-journey tries to fill out the frame and one of the big issues mid-journer used to struggle with is if you did a taller aspect ratio and talked about a person it didn't really know what to do with the rest of a taller frame and so it would often like duplicate the person's head or something to fill in that space because my journey is trying to use the frame that you tell it to fill in that space so you can see here when we started generating just basic one by one images that composition changed a little bit we either got it got a wider angle view if there's a person involved that has a lot more stuff kind of filling in this horizontal space or we got like a big warped bubbled view or we just got more of a basic scene versus This Tall spiraling image that the taller aspect ratio did so to use aspect ratio in the first place it is one of the parameters that you add at the end of a prompt so with these prompts we were doing dash dash AR Space 9 by 16 to get that taller aspect ratio so you can do pretty much any of the aspect ratios that you want now within reason like you will find some that are completely random numbers but we can pull up another reference image just to keep building on what we were doing before I found some pretty cool images on the community feed that we might want to reference so here's this one maybe I want this one to be a little taller so we're going to copy the full prompt or full command we're gonna do slash imagine we're going to paste that in there I'm going to take off the quality parameter because we're not talking about that right now and it will make it take longer but you can see here it has an aspect ratio of 8 by 11 which was you know roughly be like a standard letter piece of paper eight and a half by eleven so that is the ratio that they went with but we could generate something like do we want it to be taller or wider we kind of wanted to be taller maybe get some more of that view in there so maybe I want to go for my poster aspect ratio so if we want to do that I love that this one has no people in it and there was definitely a person in that prompt but you type in your prompt so we want City landscape art germ HD night very vague prompt overall but then we want dash dash AR 13 by 19 to make a poster that we would print on my 13 by 19 paper for my Canon Pixma whatever printer that is all you have to do for aspect ratios and then it's going to give you four images that are as close as possible to that aspect ratio again depending on whether mid Journey can compose for it or not and the and you know the reference images you use you may still end up with some black bars here or there I have had plenty of images where there's like a frame in it or something like that because it didn't really know how to make use of all that space and still wanted to give me the ratio okay as you can see here we've taken that original image prompt and generated something similar with that taller aspect ratio now because we took off the no people parameter we have a couple that primarily feature people but we also have a couple bigger ones that might have a person in it but it's closer to that mostly cityscape Focus now these are a little bit more hand-drawny than the original piece looked it had a little bit more of a painterly design but if I went in and used that original image as an image prompt with the same otherwise prompt we're getting images that are a lot closer to that style while being this taller aspect ratio that I might want to print on a poster like this would be a pretty cool poster composition with like some sort of quote here or something like that this one's a little bit more cropped in on focused on the person but it's still pretty cute little image there we got like the downtown cityscape in the background we've got the town bustling town and the foreground or the mid ground this one has the person standing on the roof which I just thought was kind of goofy and same with this one that person is clearly just floating somewhere rather than being behind the balcony but otherwise pretty cool looking images and we can do all that with the power of aspect ratio and you know control it however this also brings us to a good time to talk about your default settings for mid-journing because if you're regularly generating something in a particular way you can set some of that as defaults within mid-journey so you don't have to type it remember to type these little parameters every time so to access these in any channel where you have the mid-journey bot accessible or ndms with it you can just type slash settings and hit enter and it brings up all these buttons don't get overwhelmed there's a lot of buttons here but a lot of them are just like basic stuff so you have the different mid Journey versions that you default to I'm going to switch mine to defaulting 2v5 now there are we're going to talk about the different versions but effectively every six months or so the mid Journey team organization whatever you know has a has a new version of their image generation algorithm that does things better or differently and you can start using different ones there's also the different versions of niji Journey which is the anime version of The Journey so you can choose those instead of the default mid Journey one you have one to enable just the test whatever experiments they may have going on right now you can default to that instead of the normal mid-journey one they also have a test photo one I don't know that these are actually in like used at the moment we're gonna we're gonna try the same prompt here AR 13 by 19 and you can see here it fires off that command with those extra parameters so yeah the test and the test photo right now are not in use so it will just give me an error so we need to set that back to B5 but whenever they do new experiments for their new algorithms or whatever for their image models you can turn those on to just use it by default then you have quality so there is such a thing in mid-journey this is typically with like stable diffusion would be called like scale or the CFG um you can basically tell mid Journey how long to think about the image to generate it and so you can do high quality which is going to be slower half quality which is going to be faster in the normal quality so it uses less of your credits and things like that that you're paying for with mid-journey but then affects the quality of the final result that you can get and I think quality can go all the way up to like five or something at this point the default buttons for it just give you 0.51 and 2. and so I often keep quality two as my default just because I like you know getting the highest quality renders and ideas out there but it does mean using up more of your credits costing you more money and so on then there's stylization uh but you can control how much mid-journey stylizes your image based on its own you know creativity so to speak it's a machine it doesn't have creativity but you know what I mean like that they can inject their own stuff into it and so by default I believe everyone's set to medium which it means it's going to adhere more to your prompts rather than trying to get away from it but you can set it all the way up to very high which is going to go super stylized for everything that you do so for example I'm going to turn this on and we're going to regenerate this whole exact prompt that we just did for these taller images here slash imagine and it doesn't matter if you copy a prompt from like the website or your own stuff that already has these in here it will just it'll duplicate them but it won't affect anything so now we're generating the same image in the same aspect ratio with the same reference image but with a quality of Two and a stylization of 750 which is what it deems to be very high we're gonna set it back to medium at any time if you get freaked out if you think you mess things up you don't like where your settings are at there is a reset settings button so I can show you what these defaults are reset settings mid Journey V4 base quality medium style and then it doesn't change this bottom row at all so I'm going to set it back to high quality and V5 and then it shows you what your suffixes are here now you may have seen on some of my prompts there's a dash dash up beta I was using this controls the upscaler algorithm that it used for the older mid-journey models doesn't really apply to the modern ones so don't even worry about it then you have public and stealth mode as a paying subscriber I don't know if you need the higher tier or not you can enable stealth mode which means they're not displayed on your public profile page on the mid Journey website website as I showed you before so if I go home you can see my recent Generations but if you're if you someone else is looking at my profile you're only going to see the ones I've marked as public and so you can control that here you can enable a remix mode and that brings up the buttons I showed you here where if you make a variation either on an individual image or with the V1 through V4 buttons it gives you the prompt window back instead of just firing off that command again to generate four new images it gives you the prompt window back so you can make modifications within that variation command which is really handy allows you to remix the image and then you have fast versus relaxed mode paying users again can use relax mode which generates your images more slowly but uses up less or no credits at all I think yeah you run out you're paying for fast credits and then when you run out of fast credits you're kicked over to relax mode and you get unlimited relaxed credits you're just a lower priority under everyone else so it takes longer but it is a great way to go if you use the permutations feature to generate like bulk images while you're sleeping which we'll talk about a bit or if you just you're not in a rush you're just kind of doing this on the side while you're doing your work or watching a video or something you don't need them to spit images back to you very quickly you can toggle on relax mode to do things a little more slowly and save you some credits so here we have taken twice as long to generate the quality to and stylization of 750. and frankly we got basically the same thing like I see you can see the quality improvement but stylization wasn't super heavy over this one so those are your settings for the defaults that you can set for your suffix so you can also come in here I'm gonna set these back I was experimenting with something you can do slash suffix and it's going to say prefer suffix and then on top of your default settings here you can do dash dash AR and then say you always generate just standard widescreen images for YouTube thumbnails you can add aspect ratio 16x9 so now any image you generate if we do slash settings again you can see here we can toggle these on you know annoyingly if you do the slash prefer suffix you have to toggle your settings back but now we can combine that AR 16x9 for a widescreen aspect ratio with the V5 preference and the high quality preference and every prompt we send out by default will have these attached to it so now if I do slash imagine cyberpunk dog saving the Galaxy and hit enter I typed no parameters but it automatically filled in an aspect ratio of 16x9 version 5 and quality of two for me so that's another way you can manipulate your settings there now you can do slash prefer suffix and then just leave it blank and it's going to get rid of those and then you can go back to your settings and set those up I'm going to leave base quality on for now since we are trying to move a little quickly for this tutorial here I do wish they would just add all of the other options here in the settings menu or like a custom field or something so you didn't have to do this because there's a lot of different parameters you can manipulate but whatever and just for fun here's our cyberpunk dog saving the Galaxy these are awesome I love them speaking of parameters we're gonna we're gonna start talking about some more of those so first and foremost we're going to cover tile mid-journey can generate tileable repeatable patterns that you can then just scale infinitely this is useful for making like uh the the virtual the digital paper patterns for Etsy for backgrounds for animated backgrounds that you want to Loop infinitely there are lots of uses for it and so we just do our normal slash imagine command we'll say cyberpunk UI elements no we're gonna say CRT scanned wireframe UI elements shapes and icons pattern of UI icons and then you want dash dash tile as your parameter to make a tileable image and these days nine times out of ten I'm pretty sure you're gonna get one that's perfectly seamless for a seamless pattern every once in a while you're gonna get something that isn't and something to keep in mind is if you use some sort of post upscaler in Photoshop and you know an AI upscaler like gigapixel or something those have a chance of breaking the seamless pattern capabilities of whatever image that you got because it's going to tweak the little you know the small details on the edges and it's not going to be seamless anymore I've seen people run into this but realistically you don't need to upscale them if you're just making a pattern because the default size you're going to get out of it is fine and you're tiling It Anyway tiling just means you know repeating it in all directions so it should be fine now I wouldn't mess with aspect ratios much for tiles because you kind of want the square to be easily repeatable I can already tell these are not going to look how I want trying to get tiny little UI elements it's pretty difficult out of my journey sometimes but we can come in here and and HUD UI we can just try to find find an example of someone making icons these are pretty cool Huds but these are not what I'm looking for uh UI icons there we go you can see lots of people have just made icon packs with these and that's pretty sick we'll talk about logo design a little bit in the future wireframe icons yeah it really struggles with the wireframe design I guess I shouldn't try to focus on that here we go if we want something like this this would be a boring pattern but you get the idea yeah like these I guess they'd kind of make decent patterns yeah let's let's just go with something like more default let's say imagine I'm gonna say what are we coming up on holiday wise Fourth of July isn't really icons for that aren't like it would just be fireworks that would be kind of weird so we'll just say we'll just do Christmas cutesy Kawaii icons of Christmas symbols iconography adorable Christmas pattern design family friendly comfortable comforting dash dash tile it does great with the like cutesy Pixar sticker style all right here we go these look like they'd be perfect for wrapping paper some of them are a little goofy like they're missing a face or something so you might wanna fill those in yourself if you know in a photo editor but some of these are pretty adorable and for some uses you'd never really notice anyway I like number four the most so we're gonna splice out number four I like number one two I'm gonna go in and click that just for fun with the version 5 model upscaling is pretty much instant because it already generates these so I'm gonna go ahead we're going to go with number four again I realize there's some errors you could try to fix them yourself or just keep working on your prompt for the sake of the tutorial we're just gonna rush to it I'm going to save out this image once it loads here or maybe I'll just click the button anyway there we go so now I'm going to post a link to the description in the link to the description a link in the description to this tile seamless texture Checker which allows you to check your patterns to see if they're seamless and so we can come over here drag our image on and then you can zoom in and out and see if you see any seams it looks like it perfectly you know tiles for me and we can zoom R all the way in here it looks fine I think for wrapping paper it'd be pretty great but as a demo I wanted to show you if we run it through something like gigapixel to upscale it if you wanted you know if you're trying to brute force it to be higher resolution which I don't entirely recommend I'm not even happy with the results of that but let's just say we do that and we export it okay we have upscale it so now if we check the upscaled version with the tile checker Zoom back out a little bit this one actually still looks pretty okay I have seen people we also don't have a ton of details on the borders of the tiles if you have a bunch of like small details on the borders of the tiles you're going to start seeing a little bit of a seam I've seen a lot of people complain about it and I've seen it myself with this one we ended up okay I I it holds up a little better under scrutiny although the finer details when you print it may not look as good but this one worked out but a lot of the little details with tiling will be kind of broken by upscaling all right we have mastered a lot of the standard mid-journey prompt Generation by now you have the ability to create images in any you know size and shape that you would like you have the ability to make patterns you have the ability to use and make your own reference images to build out ideally the exact image that you want there's still some research you can do on details and we'll talk more about it but I want to take a break from The Prompt crafting again to talk about seeds you know this will this will help you take you to that next step seeds or again if if you missed it they are effectively a reference to the noise pattern that is used to generate your image so again AI art is generated by taking what's called a gaussian noise pattern which just looks like this just raw noise and trying to find the best possible route to recreate the the subject the style the composition that you have given in your prompt to produce the image that you want and it uses machine learning training to tell itself or for them to tell it when it does a good job and when it produces what you want so it can continue to make better and better results and each time it starts with a new noise pattern but you can control that and tell it to start with the same noise pattern each time now used to either by finding a image in the mid-journey feed here and reacting to it with the envelope emoji and having the bot dmu the job ID and things like that or by going to Mid Journey's website and clicking the copy button you used to be able to get the seed for every image generated by mid-journing which made it super nice to try to build something off of what someone else had done to try to get some consistent results to learn about how your prompt changes affect the result Unfortunately they have removed this for other people's images and I kind of get why in terms of like not stealing other people's images but given how mid-journey works and sources its data I can't say that's the most logical of takes to have but you can still do this for your own images so if you go to your home page on the website and find any image you have generated and you can get the seed for it so we can take for example this neon rain one we can click the three dots and as usual we can copy the the command The Prompt the job ID but we can also copy this seed now you can also get this from your images you have already generated here in Discord by just navigating to where they are on the web so I can well this one we haven't been giving a web button because we haven't upscaled yet so we can do the envelope emoji on it and then the mid Journey bot is going to give us that information including the seed that it used for the grid itself which is really powerful so we're going to take the seat and now we can use this in our prompts to better kind of Target what we were doing with this so let's say we really like the base ideas it comes up with for this cyberpunk dog but we want to tweak it some more but every time you send a new prompt even with variations it doesn't always use the same seat most the time when you send Pro I think variations sometimes do but every time you send a prompt it's going to use a new seat unless you tell it to use the same seat so say we really like version four here or this yeah this image four we just want to tweak it a little bit we're going to say make a variation of V4 and I have remix turned on for this otherwise you'd have to upscale it and then click the variation button so I'll show you what that looks like so if you click upscale then you can click make variations otherwise if you have a remix turned on in your settings as I just showed you we can click V4 here now since I want to make a variation of it I'm actually going to upscale it this way so I can use that as the image source so I'm going to say make variation I'm going to do Dash seed and paste that in there but I also want to use this as the image source as an image reference so I'm going to paste my seed here so I have it I'm going to right click copy link now I'm going to say make variation paste that in there get rid of anything after the file extension and now we want dash dash seed and I can't copy it without closing our prompt so I'm going to say stash seed space and then the number so 180270-8178 but now we want to tweak our prompt a little bit so we can continue iterating and see how our changes affect this and so I can say cyberpunk dog saving the Galaxy maybe I want to say instead of the goggles dog is wearing a VR headset with googly eyes on the VR headset now it's using the same C to use to generate this image it's using this image as the reference for the composition and colors and we're just trying to add in that VR headset element instead of the goggles so it's going to take all of that information and give us something new but it's based on the same noise pattern that it uses here so this is a way to more consistently iterate fine details that's kind of what I've been building up to over this course is I can show you the prompt manipulation I can show you the image references but then now we add in that seed element so you can just finally finely tune your prompt to continue generating something different each time and that is the AI art generation process as the user of just tweaking more and more of those parameters to try to get what you want you can see here already it's not done yet but it's not doing a VR headset because of the image prompt here it is sticking to those goggles so again we go to our old classic photo bashing technique this time I'm going to use Affinity photo which I've talked about the Affinity Suite on my channel before I have a video on the version 2 Linked In the description it is a much more affordable alternative to Adobe image editing software and they have a much better business model in terms of buying once and not subscribing and it's super affordable so I'm going to make an image we're gonna make it 1920x1080 that's fine we're going to take our original reference image here paste it in that gave me text but it also gave me the size of the image that is super weird there we go I don't care about the text we just want the image we'll hide that text layer and now we need to find an image of a VR headset just go to classic either Yandex image search or Google image search and we want it facing the same direction or to manipulate it in that direction thankfully that's the cliche pose that everyone uses anyway if you're not comfortable with like ripping images from Google Images even though we're not using this in our final product it's just to you know build a reference image like you'd send to any artist you can use a stock Library so I am subscribed to envato elements they are not a sponsor but I do have an affiliate link with them if you want to subscribe it's just a subscription service to get tons of stock assets I use in almost every project I do video or image wise I'll have an affiliate Linked In the description that will give a small Kickback it's for what it is for like a year subscription it's super affordable but I'm gonna just search VR headset go in here to the photos that photo is about what we want so I can go ahead and just license that download it whatever that's fine find it drag it into our Affinity comp here if you want Affinity photo tutorials to do some of these Basics let me know they're not going to perform as well but I would be happy to show you all these Basics it's a little out of the scope of a video specifically about AI art but I can walk you through some of the basics so I'm just resizing my image so that the headset fits over to oh gosh darn it all right can I lock yeah I can lock the dock we don't need to manipulate the dog anymore so that layer is locked I don't like how Affinity is like layer selection and stuff works compared to photoshop there's a lot I don't like about a finish Affinity having used Photoshop for a very long time but I make it work all right so to see where we want the mask to go we're going to lower the layer opacity so we see a blend of them so we can line it up about where we want that is about perfect and then what we're going to do is we're going to create a layer mask that basically just hides everything except for the VR headset so I'm going to click this button right here which creates a layer mask we are going to use the brush and I don't have my tablet plugged in so we'll do this on the mouse we need the black so with a layer mask White shows through black hides and so we want to use the black brush that is the selection brush that is not the paintbrush I'm going to use the black brush to hide this and this is basically erasing but the advantage of using a layer mask instead of erasing is that you can more easily refine it and continue like showing and hiding over time versus having to just control z a bunch to undo your erase it's a much finer more professional way to handle this that makes your life a lot easier in the long run so we're just masking away the underlying or the this overlay layer here to where we just have the VR headset we don't want the guy we don't want the background we want the headset over top the snoof he's a good boy he's saving the Galaxy but he needs to be doing so in a VR headset trying to decide if I want the ear cuffs or not so just as an example since I did talk about this say I just accidentally drug through there obviously I can control Z let's say I'm already making other adjustments I can hit X and swap the colors from black to white and now white painting white on shows that back through like I said just makes your life a little easier to refine these selections once you use this workflow a bunch you get used to the like back and forth of it it makes things so much easier so there we have masked on the VR headset so that's generally what we're going for but we also want the googly eyes on there and instead of just since we're already doing all of this instead of just hoping mid-journing comes up with the googly eyes we could also pull up an image of some googly eyes so this I am going to use Google images for just because it's easier all right we're just going to use this one again this is a reference image I'm not making something out of it oh and this one did actually have transparency even though when I dragged it it didn't look like it all right so we're gonna do a little something like this there we go so obviously this is not the final image that we want overall because it's just a poorly done Photoshop and we could spend the time to Affinity edit I guess not Photoshop we can spend the time to color grade and all of that we don't really need that it's fine as is we're going to export as a PNG again to our scrap folder we made before which I recommend keeping in your organization we're going to say cyberpunk dog VR googly eyes let that save out we're going to go to our reference images Discord Channel we're going to drag that one in copy image link and we're going to do the same thing we did just a second ago with that remix so now that we've made that image we can click make variations again paste in that image link into our modified prompt cyberpunk dog comma dog is wearing a VR headset with googly eyes on the VR headset saving the Galaxy we may need to put that part at the end but then we still do Dash gas seed 180270-8178 and submit okay anytime I try to make a variance I've done this like a million times anytime I try to make a variation on the upscale it sticks too closely to those original goggles even with our modified image so what I would do in this case is try it on the V4 itself so we're also going to try dog is wearing a VR headset with googly eyes that's just seed1802708178 try it on that and otherwise just try the original prompt itself without that original variation at all specified so we just copy what we typed without remixing at all and we'll try that too slash imagine so we have a reference image we have the original prompt that we have modified to include the VR headset because you see this one is still trying to stick to those goggles whereas if we feed it something new still with the same seed but without that original reference of the goggles at all we should get closer this is all about experimentation I've had to do this way too many times to try to get rid of those goggles it's what the process is all about yeah still those damn goggles with the variation here is our from scratch generation happening and you can see here it's already doing better we still got a goggly look to it I think it's because the eye is on the VR headset kind of look like goggle bits but we're getting closer maybe if we took the eyes off entirely it would be fine or perhaps we do it all without a seed like that that this also shows the power of sticking to the same seed maybe that just adheres too strongly to that original design because if we scroll back up to that original image they all have goggles on except for that one but they all have goggles on that one does get closer like that that that one is definitely more of a VR headset um so this time if we just generate entirely without the seed we'll compare because this one still sticks to that original design these look awesome like that fourth one is incredible second one looks a lot more like a VR headset the third one looks like some sort of combat Mech headset but they're all sticking to that original look that we had here so if we generate a new one without the seed we can compare it to the exact same prompt with the seed clearly mid-journey has no clue what to do with googly eyes but these are the cyberpunk dogs wearing something that looks a lot more like a VR headset now so we're we're making a lot of progress here so you can see here if we use the seed it tries to adhere to goggles a little more without it we get a lot more of a headset design so that's the power of seeds I honestly don't use them a ton but if you are trying to iterate minute details you're gonna want to stick with them if you don't want to manipulate fine details and stick with your original path and you want mid-journey to make a lot of those decisions for you to help kind of find your way that is where the stylize and the chaos parameters come in so thus far we've been building up your workflow of trying to like hone in on a specific image idea that you may have what if you wanted to add Randomness to it to have it make those decisions for you or you can use that stylize and Chaos parameter so we're going to do imagine we're going to take paste in our prompt without the seed but this time we're going to do I'm going to take off the quality too because I want it to go quicker we're going to do dash dash s which is stylized you could also just do style and then you can go with at least with version five you can go from zero to a thousand and as we showed in the settings part the default is going to be somewhere around 500 already for the medium stylization it might be around 250. I don't remember off top my head but it's already doing some but you can inject some more in there so we can say dash dash s and actually we are going to use a seed here so that we can continue building upon this image so we're basing it on this generation we're using the same seed the same prompt same reference image all of that but we're going to add in a stylization just full fat 1000. and then we're gonna do the same thing and we're going to do 750 and we're going to do the same thing and we're gonna do 200 and we're going to do the same thing into zero we're gonna let all those process okay now we can watch the progression a zero stylization has a very mid-journey style as people often refer to it so it's lacking in a lot of the detail it still has some photorealism but like it's not doing a whole lot it's just kind of some elements slapped together they still look kind of neat it's kind of what you'd get out of dolly or adobe Firefly or something where they haven't trained its own style and trained on itself a whole lot yet that's what zero stylization at all with 200 stylization we're already close to our original prompt which makes me think maybe two 200 250 is the medium option by default because this very much looks like our previous generation that we had we get a lot more detail on the dog we get more stylization of the headsets we get some background elements coming in some weird particles on this one weird choice but it did it then looking at 750 we get a lot more weird designs in the headsets this dog's wearing headphones instead of a VR headset but its eyes are really bulgy we get some more of that kind of injected in there it's a little more just with stuff with backgrounds but I think looking at these smaller details are interesting you get to see there's just there's just more weirdness to it as it goes because it's trying to like really hone in on whatever cyberpunk style and then 1000 mostly kind of just looks like 750. however there are some easy to reference uh Resources with regards to seeing direct differences of the impact of stylization and chaos and things like that which we'll talk about after we talk about what chaos is so that was stylized where it just tries to you know we start with a very kind of plain looking thing develop into more of a stylized look over time chaos adds in Randomness so let's say we really like 750. we like everything we have here we're going to copy this prompt as we had before same seed same everything we still have the 750 stylize but this time we want to add chaos chaos again it's just Randomness the it's basically giving mid-journey more leeway to vary away from your prompt to do something that it thinks might be good just based on wherever it ends up chaos is basically a percentage from 0 to 100 so again I'm gonna fire off the same set of jobs here all right so chaos of zero is of course what we just generated the first time the image looks basically identical to what we had before so that's a good starting point for reference and then moving up to a Chaos level of 25 you can see here the dog type start changing we start getting some different elements we start getting some black bars on some certain images which is frustrating because it stops adhering to your composition desires or like references as much the goggles on this dog just go absolutely wild and that's only with a 25 Chaos level with 75 with 75 things are pretty much already out the window we get this really I love I'm in love with number three here the style on this is incredible with the the gradient on the goggles I wish it aligned the same but I could probably tweak that in post this one again wrong aspect ratio this one's just a complete that looks like a rugby helmet we get some wild stuff and then full 100 uh chaos it's very similar but just you'll probably get the primary subject you asked for or you your image reference has but very little else that is what chaos does like it is wild and if you wanna if you want to go real crazy with like mid-journey prompts just come up with something vague and then use Max stylization and Chaos so this works best on V4 I'm going to demo it here on V5 V5 tends to be a little bit more literal and like hardcore attending to the words you say whereas V4 and older tend to get a little more abstract but we're going to try to push V5 to be more abstract so we're going to say the desperation when sleep just won't start something edgy I guess Dash Dash AR two to one for something super cinematic dash dash s for stylize we'll even say stylize 1000 chaos 100. you can use dash dash s or stylize for the stylization percentage okay four fine I'm four fine I went ahead and did a V5 and a V4 one so with V4 you get something a little more abstract a little less photorealistic more in that mid Journey style but it just goes crazy it just takes whatever it takes from your words and does something so you've got that sense of desperation and a sense of sleep a kid's yawning that one's just pouting on what is a very sad excuse for a bad that person's just really sad in the middle of their bed same with that one both of them are just like coming out of the middle of their bed somehow with their torso and then that person's sleeping up against the wall but just very like you know super crazy and then V5 defaults to a very photorealistic Style and so we've got two people trying to sleep on a pillow which makes sense that guy looks sad she looks sad pondering the time because it's I guess 5 25 a.m and she's still not sleeping and then that one's just a clock outside producing its own Spotlight because why not you can go crazy with like super abstract prompts and things like that like if we just do a single word but you can also use these with the permutations feature which allows you to generate a bunch of different images at once from your prompts uh to kind of find a path to go down as a way of like finding your inspiration or finding the next step for what you need to do we'll be getting there in a little bit permutations is a really awesome feature all right now this one took forever to generate for some reason but I just fed it one word wonderful with a Max stylized and Max chaos rating and it gave four completely different images we got a random anime character which I guess it might be wonderfully drawn but I don't know why it would exp you know invoke the emotion of wonderful or the feeling of wonderful the adjective of it you've got a happy couple jokes yeah checks out dude kind of looks like I don't remember his name there's an actor that that dude kind of looks like there's a cute little puppy drawing I guess that's wonderful and then like a swirly Vortex which could be like a wonderful like wonder of the world kind of thing but you see the complete Randomness you can get with this kind of thing but maybe you're just looking for something to do and you just feel feed it a single feeling or word you see what it comes up with and then you start iterating from there and you know maybe I want to do V4 and I want to be like wonderful Vortex of emotions in hyperspace and at that point you could turn down the chaos in the cylase if you want or you could just still let it kind of do its own thing and probably take forever to generate when you're on the max setting like this oh it's actually generating something pretty similar very cool [Music] now up to this point I have given you the understanding or helped you build the understanding I guess to build your own prompts to reference your own images to even start photobashing your own reference images to create the images that you want but there are other resources that can help you fine-tune this to the nth degree because there are so many different styles and and graphical elements and ideas that you can incorporate into your prompts that you're never going to think of because most of them are like made up words and things you would have never thought of but there are resources to figure it out the first one I'm going to point you to is this uh ebook called the AI art Master guide uh someone reached out to me about this and I checked it out and it's actually really handy this one costs money a fair bit of change but if you want something that's like an all-in-one easy to reference Resource as a lot of people have asked me to make and I frankly just don't have the time for right now this is a really great resource that they're going to keep updating over time for any changes to Mid journey and things like that both to get started with mid-journey itself but also to reference some Styles which is mainly what I feel are the like main advantages here because there's all all of these Punk and core and wave and those those kinds of like style things that you'd never think of and so there's the obvious one like synthwave cyberpunk but then there's like Forest core hiking Punk those kinds of things so this one has sword punk as a city light core neon Punk Cloud Punk like so many of these weird combinations of things are styles that mid Journey has found and scraped and identified goth Punk City Deathcore doodle core like this looks incredible I would have never thought to include any of these words because they're not styles that I reference on my own obviously some amount of art studying and knowledge and you know the the study of art styles and all of that is going to benefit your ability to manipulate an AI art generator but some of it is just knowing these like Niche cultural things to achieve the styles that you want and so it has libraries with comparison cubes of all the different styles and things you could reference to help better inform you and this is what I was talking about before of having something that more cleanly kind of shows you the differences here this is a really valuable resource in my opinion like look at space solar Punk like that looks incredible so it has stuff like that it's got um Advanced like just general prompting tips things like that like I said this one costs a little bit of money I will have an affiliate Link in the description that gives me a small Kickback just so you can check it out for yourself like I said it's really really powerful but there are some free resources you can use too I'm not going to shy away from them the first one is mid Library dot IO which has tons of different like Styles film types you can references photographers and artists if you want to go with names again I recommend sticking away from it a little bit but it has tons of stuff baked in here and you can just search for it so like I can search for glitch what kind of glitch art stuff do we have glitch art we've got oh there's like blog posts about it and stories and things like that but if we just want to look at genres and art movements we've got glitch art and V5 and V4 and easy Journey and then it starts pulling up some sort of similar reference ones I I would say that a lot of the Styles and style differences are going to this one won't have as much as the PDF but it does have some cool stuff you know it depends on with all of these it's going to depend on what style Direction you're going in because plenty of people focus on like specific things but this one has tons of you know 80s fashion a different style of paintings and all the stuff so this one is like super useful in that regard of seeing comparisons of different styles that you might want to pull from you've got everyone references in the style of Akira you've got that going another great resource is this GitHub page for I referenced this in my original course for Mid Journey Styles and references this one is less complete compared to the previous two resources I've shown but it does stay updated and it has these cool buttons to navigate different things so like for example you can add in details about like effects and shaders into your prompts so you can click to navigate these and so there are different types of Reflections you can reference and it has comparisons for those screen space Reflections diffraction grading reflection and a puddle you've got chromatic sfx and it has all these comparisons for this so it has like tons of different ideas that you can kind of navigate in the buttons here as well as different prompt parameters to play with which is pretty cool you got geometry references you can make so you can say like 4D hyperships and get crazy designs from that a hyperplane like there are so many weird things that you can combine in mid-journey that these references provide to you and then the last one in terms of websites is just the official mid-journey docs page docs.midjourney.com docs it has how to get set up how to do different things and there are some user guides with regards to like commands and certain style things that if you just need help figuring out how to use my journey you can use my last last free resource is actually in the mid-journey server if you go to the mid Journey server which you have to be a part of to use mid-journey well I guess that's you could be in another yeah you have to be especially now that it's not free they have I know it's hard to find anything in here they have specific chats based on like prompt crafting and getting ideas and seeing people's final images and stuff but they also have in here it's under Community forums there's a prompt fax one and you can search here for like different ideas like the consistent character thing you can search for character tips and it's going to find new threads with that um it's ideal not to ask questions here but you can see like because a lot of these are marked as help requests but the ones that are marked as tutorials with the flags and you can actually like search based on these categories so there's challenges there's libraries of information or tutorials if you click tutorial you can see tutorials on how to photobash how to do magical effects and they'll walk you through some of the prompt tips for that lots of cool stuff there so you can get lots of ideas and that's where a lot of my ideas come from that some of you are like mind blown by they just come from you know all the contributions here to build together a big Community understanding of how to use mid-journey now that we've covered or at least given you rabbit holes to go down to get the resources for everything you possibly need to know to make like normal prompts and and all of this with mid-journey it's time to get into the more nitty-gritty so you can really control mid Journey the way that you want and we're going to talk about the different versions of mid-journey so again if you pop up your settings menu by typing slash settings you have effectively nine different versions of mid journey two of them don't work so two of them don't work so seven versions of mid-journey that you can work with mid Journey one through five four being the current default um V5 is still like being finalized and it's a more photorealistic one and it's one I recommend using just moving forward because it's where things are going but if you want more abstract stuff go with before we'll talk about that in a sec I'm getting ahead of myself then you have two versions of the niji journey which is the anime model of mid journey and then you have two of the tests versions that currently don't do anything but sometimes they'll turn on experiments and they will do something and they'll post announcements and things like that V1 through V5 is just mid Journey as an organization training their image model on the results that they got from mid Journey itself using the rankings on the website and things like that to com continue to improve the accuracy of the prompt generator you know the prompt responsiveness as well as to do whatever magic they do behind the scenes to manipulate your prompt into getting the result that you want without being hyper literal about your words V1 was super abstract painterly kind of gritty looking V2 was still there V3 was getting better on that produced a little bit more beautiful at a glance but lots of issues images V4 is where it really started to get great looking but it still struggled like prior to V4 people just looked horrible people weren't super useful at all but V4 still struggles a lot with hands V5 is a lot better with hands and overall is a lot more photorealistic by default you can still do non-photorealistic Styles but by default it's set to operate in a photorealistic manner and V5 is kind of where things are going with it so you might want to stick with it but you can choose any of these other versions as you see fit to continue iterating in more abstract or painterly or whatever styles that you want so you can experiment with running your same prompt through the different versions and see which ones you like just to see you know how you like it and maybe maybe you want to experiment with some of the older stuff a lot of people had a lot of fun feeding abstract phrases or song lyrics or something in mid Journey V2 and V3 because it was so abstract and non-specific it could do weird stuff like that throughout this entire course so far we've been covering V4 and V5 like that's what we've been using but there's also the niji journey which is the anime version so now with this which you can use there is a whole dedicated server for an EG Journey here that you can join similar to the mid Journey one um and then add the bot and you know interact with it separately but you can also use normal mid-journey with the dash dash knee G5 or 4 to interact or you know to generate your prompts with that model so now we can create anime style images so let's say we go back for example yeah we'll just take our prompt and most of the prompt parameters and everything is going to be the same so we can take our prompt of our cyberpunk dog do slash imagine and when you add the niji journey bot to your server you'll get a separate imagine command for that bot but we're just going to use the normal one since we're using the parameter from an easy Journey just to keep it simple here we're going to paste in our prompt we're going to get rid of the seed part we don't want that right now so cyberpunk dog saving Galaxy VR headset aspect ratio 16x9 when we hit enter it's going to send it with the niji 5 parameter by default because we added it in our settings here and you can hide these messages at any time with dismissed message and it's going to render that with the niji journey style now that that style that model doesn't always stay anime so there is a chance by feeding it a photorealistic image like we are here it's still going to produce something closer to photorealism than anime if so we need to get rid of our image source and kind of start over a little bit or perhaps apply a filter to our reference image we built that makes it more line Arty anime style to try to compensate for that let's see what it comes up with all right I went ahead and sent it off without the image reference as well but you can see here this is with the image reference so it still has very similar looking images but it did convert it to a more anime style and take some hints from that we've got more you know anime style textures and drawing here it gets a little abstract in a couple of them but it does a good job it also confuses the aspect ratio these are all different aspect ratios with black bars than the one that we wanted it can happen sometimes there's you gotta fight with it a little bit but sending it off without any image reference at all you can see here we're getting something completely different though it's doing a better job of keeping that VR headset in here and this it has a lot more of that really bold simple color anime style that you might see from stuff but these are these are pretty cool like it definitely does I don't know what happened here it combined a person and a dog but like doing a pretty good job so you can generate anime style images with niji journey using all of the same everything you've learned about mid-journey to this point but specifically catered towards anime instead of you know photorealism or painting or whatever now we can take this a step further as the niji journey model has a dash dash style parameter with I believe at the moment just two options that you could add into it so we're going to take I guess we're going to take this new one we're going to use the same seed so I'm going to envelope react get me the seed for this prompt we're going to do slash imagine cyberpunk dog saving the Galaxy Dash seed the seed for what we just generated but we're also going to do Dash Dash style expressive so expressive is better for character portraits to try to get more oh our aspect ratio got deleted expressive is more built for character portraits to try to get more expressive you know facial portrayals and things like that out of it you can apply these styles to anything you're just going to you know they're catered for specific ideas there is expressive and then there's Scenic which is built more for getting wallpapers out of mid-journey to get more of an anime style wallpaper now we can try it with the same command here just to see what it says um but I have seen a lot of people on the niji journey server doing style Scenic with like portraits and stuff and it doesn't really do a whole lot in that regard we can look at a comparison real quick but it would if you're building a oh it just failed to send my command at all I have broken mid-journey there we go so that is our expressive one so we've got the puppy dog is smiling that one's grumpy that one's panting that one's super angry laser eyes mode we got more expressions and more like Avatar kind of generation with expressive enabled ah that one failed it was just going oh no it's still going mid Journey's having some issues here always happens when I go to film tutorials it's a very popular service at this point so I guess it stands the reason that they're gonna have some errors all right with Scenic enabled we still got the smiles and stuff but we have you know it has more of that beautiful background element built in to act better as a wallpaper so you can still use Scenic with portraits and other subjects and it build more of a wallpapery kind of thing but it's more built for wallpaper so for that I'm actually going to reference an older prompt we did I'm going to scroll all the way up here do this little yeah this this one the little landscape we had going on here I'm gonna use the one that we did with the image reference so I'm going to take this prompt I'm going to make it widescreen since I want it to be a wallpaper we're gonna do this where we have the the art germ landscape that we did before in the kind of Miyazaki style and instead of a poster aspect ratio we're gonna do 21 by nine aspect ratio so by the way at least last time I tried it you cannot do decimal places in your aspect ratio so like cinemascope would be 2.35 to 1. you can't do decimal places in the aspect ratio so you would need to convert it up to a number that doesn't have a decimal place which in this case would be 21 by nine so we're going to fire this off with niji journey oh but I didn't do Scenic so we're going to do the same thing I'm gonna say Dash style scenic we're going to let those process all right now here we definitely got more of that wallpapery aesthetic it moved the people off to the side tried to focus on the city we did encounter finally that issue with the aspect ratios that I was talking about earlier we used an image reference or a reference image that was vertically oriented for a widescreen generation and it put a bunch of black bars in it because it didn't know what to do that it did that for this normal one and for the scenic style and that is a problem and so again our options are we could crop it or like stretch it out or something or just not use that particular image for Our Generation I'm gonna go ahead and fire this one off without the reference image at all but you get what I'm saying like it's it's a pretty cool little feature that you can do to start generating anime stuff alongside your normal generations and maybe find a style you didn't know you liked before if you're not interested in anime at all you'd never have to touch it that's totally fine and now we've generated some super wide it's still struggles to adhere exactly to the aspect ratio that I specified but we could crop or whatever a little bit but some pretty cool anime style wallpapers that a lot of people would be happy with they're not perfect but they're pretty cool now what if you have an image that you see from someone else for example I have my entire inspiration section of my server I really like this image but I want to know how to make it how how do I how do I how do I get a prompt from an image well mid-journey now has a describe feature that allows you to generate prompts from an image so you can get closer to your idea and you can use this either by sketching out or photobashing your own image or by just taking an image and using slash describe on it so we're going to start with our reference image here of the doggo with the VR headset let's see if we can get closer to this image with whatever mid Journey tells us so actually I'm going to stay in my reference images chat because the slash describe messages unfortunately disappear after a while whereas your normal image generation messages do not it's weird so I did Slash describe and I hit Ctrl V to paste the image otherwise you could drag and drop an image on it this works on mobile too and send that off and it's going to present us with four prompts that resemble that image we're trying to generate it also has a secondary benefit of at least temporarily while this message is here it gives you a link you can copy for that image without you having to manually upload it but again these messages don't stick around but here we have the dog wears a virtual reality headset with blue eyes and the style of rendered Cinema 4D dark Azure and purple quirky character designs McDonald Punk the cool thing about describe is it will show you so many ways that mid-journey problems work and so many random phrases and descriptors that you would have never thought of that mid-journey things to use that you can then incorporate into your prompt craft and so you're given these four prompts and you can either manipulate them or do whatever you want or regenerate prompts or you can just go ahead and fire them off so since this is my reference images Channel I'm going to do this over here in the other one we're gonna we're just gonna fire all these off so I truly believe that the most powerful way to get your you know to build towards the image you're trying to achieve is to photobash your own source images or Draw them if that if you're capable of that did we do three yeah we just did three so we need four and then using describe and blending which we'll talk about in a moment and reference images to build towards your desired image however you can also take images and do this so we're doing this with the reference image that we photobatched from a mid-journey image and it's just some other assets that we found and we're already getting you know results that are going to resemble what we want here all right so these are the problems that it came up with based on the suggested prompts from using that describe feature so we're getting things closer to what we kind of wanted this one doesn't have virtual reality headsets at all but it's because it says it's a virtual reality dog this one definitely has a dog and a headset we've got ones that are close to it and then one of the prompts of the four that it suggests usually goes off in some weird directions so these are like Funko Pop Style robot creepy dogs or something can't necessarily use that but what we can do is start modifying the prompts to include aspects of each other that we want to include as well as just use the reference image as a reference image in that prompt because we didn't do that by default by generating these and so then what I went ahead and did is paste it in that reference image that we used combined some of them so I really liked that this first one specified the dog wears a virtual reality headset so I specified that and that I took out the dog in Hood scene and just put in cyberpunk otherwise Let It Go and we very quickly wound back up at that original dog grid that we were achieving in a different way and I do think the googly eyes even when we don't specify them kind of tell mid-journey that it has goggles instead of a virtual reality headset which is which is fine we could we could take those off but you get what we're going for here we've combined our reference image everything that we've learned about prompt crafting and the suggested prompts that mid-journey gave us together into one super prompt to really fine-tune things and we wound up back at a similar point so it's very interesting to look at the different paths you can take to get to the same final piece of art if that makes any sense now just to show you that you can do this with other images and stuff I had this very cool like old magazine style sci-fi collage I found on Instagram ran it through describe got some prompts so I chose I think the first one that I wanted to reference combined it with the original image as a reference image and we're getting something fairly similar I would have to do some photo editing some color grading styles to really kind of get the final look that I wanted but some of these are really nice like I really like that second one and that third one there so I could upscale those and play with those for the future but you can combine your prompts combine your describe prompts and build out this ultimate prompt that can really help you fine-tune where you're going now if you don't have a reference image and you just have an idea in your head and you want to go based on that and want ideas for your prompts there is another free tool you can use for that as well this one is from a website called noonshot and it's just a mid Journey prompt helper I've shown a couple others before and they ended up going paid access so I'm trying to continue showing free ones but we can specify that we want they weirdly don't have version five up here you can upload a source image as an inspiration image which is pretty cool but then you can just start you know typing your main idea so what if I want a Sci-Fi desert landscape with planets in space in the background obviously that gives us our main prompt but then we can start searching for other things so styles and start looking at different styles that we want to include 1980s absolutely this little lightning bolt is going to be your waiting which we haven't talked about yet but how much it should be emphasized so we're going to say okay we want that style continue now what about lighting Consular lighting accent lighting crepuscular rays which are just God Rays direct sunlight blinding light got all these different styles you can work with and choose from or search or whatever we're gonna say fluorescent continue camera all different kinds of options here as well obviously this isn't going to have everything that mid Journey supports but it gets you down the path of trying to figure it out so DSLR we'll turn that up a bit that's fine we're not going to do artists colors neon blue neon purple RGB depth of field shallow depth of field with focal blur you can have your quality options here your stylized options your chaos your seeds all of those normal things materials if you want to choose like a material of your subject we're not going to do that because the kind of scene we're doing and then your size your aspect ratio so we can do two to one I don't know why their options are so limited here but you can manually override it yourself you're not you know restricted to these you can also specify things you want to avoid so trees text buildings and that adds the dash dash no parameter to avoid those things this doesn't always work but sometimes it does so it's worth trying anyway so we're going to copy the prompt and come over here back to our Channel imagine and I'm going to swap out V4 for ve5 and actually instead of two two by one I want one by two so I can make it tall like a movie poster fire that off now this is a free tool available to you for prompt helping you can combine that with the other resources the docs pages to get all of your ideas together and just kind of take cues from this but where the other ones are you know lots of information you have to go through and like still build the prompt yourself this creates The Prompt for you which helps a lot of people so you see here the prompted created is what I typed sci-fi desert landscape with planets and space in the background but then it added in 1980s style with a big weight we'll talk about image weights in a moment fluorescent with the weight DSLR with the weight neon blue color neon purple color RGB depth of field has very much gotten away from our landscape idea but that's because all the other waiting was put on this and we'll SO waiting is the ability to manipulate your prompt based on telling it which elements are more or less important and so you do that with the colons and the the numbers here so you can see here I told it that 1980s was very important was the most important element of all of this DSLR camera shooting was a little bit more important and then everything else was one or not mentioned and so it built the image based on that parameter but if we come over here we copy the same prompt and we put a image weight of five on this first one and then maybe for neon blue color we want negative one as in it's we want no neon blue at all it will take that in mind and it will now affect things based on that waiting and you can do the same thing for image weights and that's with a different parameter so if we take this image source and we fire off the command we just fired off we can also do dash dash IW and you can go all the way from 0.5 which is going to tell it to treat it as the least important element in the prompt but still to consider it all the way up to two so we're going to do image weight of two so it's going to heavily consider our image weight or our image reference image alongside all of the other weighting that we did here you can see here this image where we provided more way more weight to it being a desert landscape with you know space in the background as a five above everything else adhered much more towards what we were aiming for Than This original image which went off in nowhere because it was mostly focused on the 80s and the neon design so that is very important to consider and then with our image weight added in for a reference image it now generated all the ones it aired out on but you can see here it generated something a lot more close to our reference image in the first place because we put so much weight on it now similarly if we fire off the same prompt but we set an image weight of 0.5 it will instead focus more on the text so IW 0.5 and then we can change this one as well we can say 2. we can say 1.5 for 1980s give that a second there it still considers the image reference the reference image so here it still considers the reference image we fed it but it then you know tries to do its own thing so we get a couple with actual you know a person in it we get this fourth one which is incredible number one Harkens back to that idea I shared before that if you have too tall of an aspect ratio and it doesn't know what to do with that composition it'll just like repeat images so we actually have two square images in the one image because it just didn't know how to fill the space so pretty cool stuff there next I want to show you how to iterate more quickly by just generating a bunch of different images this is best done on the relax mode the slower mode so you don't waste all your credits because it can get expensive fast but it can help you kind of find your path on different things you want to do the first is using repeat repeat is not super useful it's a parameter it's not super useful unless you're using some sort of randomization element so we are going to take our prompt that we gave mid-journey here but we're gonna we're gonna take this prompt we're going to say slash imagine and we're going to add dash dash repeat at the end but it's just going to basically regenerate very similar looking images over and over and over and while that might get you one percent of the way towards the next step that you want it mostly will be a waste but what you can combine it with is with chaos so we can do dash dash chaos and say like 50 and add in Randomness each time and then we need a number for the repeat so let's say we want to repeat it three times now it's going to generate the same grid of images three times but with a chaos element of 50 so each iteration will be a little bit more random it's going to ask you are your shirt because again it's gonna it's gonna spin some credits or we can tell it to fire off and then we absolutely have to wait for this for a moment we can show the template which is just going to show all three at once or we could have just waited or we're gonna wait for these jobs to show up because obviously it's waiting to start I I think mid-journey's kind of tired of me generating so much at once here and here we have three new grids based on our prompt based on our image prompt all of that but with some Randomness to it so we've got completely different results that are at least somewhat similar to what we want this first one's cool this third one's cool in prompt number two fourth one is cool this third one is a completely different direction but interesting that's like a lava World on number one uh number two is amazing here number one is pretty stellar number three is pretty awesome as like a photo replacement so that's how you could use the repeat command to add in some some chaos but if you want to specify the things that you want to show the differences of that's what permutations are for you can now use the squiggly brackets to do permutations which generate different results based on the options you specify so I'm going to show you this in practice because it's kind of hard to explain without seeing it so we're not going to do the repeat or chaos this time we're going to do slash imagine we're going to paste in our prompt but for example for Image Weight maybe we want to see 0.5 1 and 2. and then when we fire this off it's going to generate three different grids with those different image weights you can also do it with different aspect ratios you can also do it with different descriptors in the first place so we're going to fire this one off it's going to again ask you if that's what you want to do we're going to say yes let it do its thing but we're also going to do one that doesn't tweak any of that we're going to say image weight of two but then we're going to replace desert in the desert landscape we're going to say desert forest magical and post-apocalyptic building landscape and now it's going to generate one two three four grids based on each of those ideas inserted and then you can do permutations of different things all at once and just send out a ton of grids you just gotta stay organized and make sure you're keeping track of them because it can get overwhelming very quickly but we're going to send that off it's going to be like are you sure you want to do all those four prompts and I'm like yes I am boy we're gonna send them off okay I made a fatal mistake in my permutations I still had image weight of two on So it's actually not going to do a great job of showing those different landscapes because of that image weight of two so I'm going to do that again and paste that in there we're going to take out the image weight entirely and now we're gonna do it desert Forest underwater and magical but you can see here if we scroll up to our first permutation option we've got an image weight of two which very strongly has that image weight still ends up adding in some people pretty sick shots honestly but like these are incredible this is an image weight of one which gets further away from our source image that one's just plain dumb but that's fine and then image weight of 0.5 which goes back to that previous grid we saw where the images are just not really related to our image weight at your reference image that much we do see some influence like this of the desert versus magical versus Forest landscape and this one with the image weight of two where like that one's clearly a desert that one has some more magical like Vibes to it but we're not getting the forest or anything because of that image weight so we're gonna wait on these new four prompts to generate all right so with our updated permutation here we're getting a little bit closer to the 4sd landscape going on over here this one's got a little bit more color for the supposed more magical landscape since we're using so many details from our original prompt along with the reference image it can only get so far away from it um but that you get the idea of what permutations can do here if you use them carefully and use them in different ways in your process it can be very helpful I also fired off one where you can compare like all you have to do is use the little squigglies for whatever things you're separating so I also fired off one where I wanted to see what one of our favorite prompts from this looks like in the different versions of mid-journey and so to do that you would copy your prompt slash imagine and where it says your version number you do the squiggles and then one comma two comma three comma four comma five squiggle and then you fire it off and it's gonna ask if you want to do that and that's what's happening here now you will notice it to give me an error version 4 does not support image waiting for your reference images so it's only doing one two three and five but we are getting those options rendered out here so this is with version five of mid-journey no this is with version one and mid Journey it still has the V5 at the end from my settings but that doesn't actually change anything so this is version one of my journey you can see the style difference I was talking about where it's just very like Scribble scrabbley this is version three which is getting closer to photoreal but still has that weird glitchy painterly Style version five of course looks a lot more photorealistic and artsy and version two is in between version three and version one as you can imagine the next skill that you can use to improve your Advanced prompting and image generation here is blending images together this can work I have a whole video on this if you want a deeper deeper dive on it this can work in some situations but often it could provide results that you don't necessarily like but you can use that blending to go even further so let's take some some cues from my inspiration here yeah something like this and you will need the links to it you can blend any images and you you could just generate it by doing slash imagine and then putting them in there and some details and not putting any text just using the reference image links and that would be okay but it works a little bit more smoothly if you use the blend command and then you copy paste the images themselves and you can go up to like six I think or something like that so we're gonna come in here to one of my inspo folders copy image this will be number one come back down we want to combine it with maybe that space image copy image this will be number two uh it's gonna be annoying about it blend image one image two and just send that off and it's going to take the style the colors the composition of these two images and try to combine them now if you have different aspect ratios as I do now you may once again end up with black bars or stretched aspect ratios because it doesn't like interacting across aspect ratios honestly that's a pretty cool blending of those two styles of that glitchy chromatic effect with the grids and the Sci-Fi desert landscape on its own that is pretty cool but if you're perhaps not super stoked on some of the smaller details we can then take this a step further with the describe feature so let's say we really like number two we're gonna upscale that one we're gonna copy the image we're going to do slash describe on it we're going to see the prompts that mid Journey thinks would create that image and we can continue to build on that so now we have these prompts that we could work with to try to generate and see what it comes up with in comparison to this and once again we could combine that with the next step of using this as the reference image in our prompt that we generated from the describe feature of the Blended images you see how you can build all of these features into different layers to keep building out your prompt so we're gonna see what this looks like super quick you can already see before it's even done this version of The Prompt that mid Journey suggested is very focused on this person or as in this image the person's kind of off to the side so we could take that composition as a reference feed that as our reference image take that prompt that we just got from describe and now feed that through and it's going to take notes from our new composition or from our older composition for the new image this one is still pretty cool it's definitely doing something really cool but let's see how it handles using the image reference this is one of those instances I mentioned was a thing it says open on website for full quality the compressed version on Discord it knows is going to be less than what majority can actually deliver and so then you have to go back to your mid journey home page and wait for it to show up refresh your feed and find the grid there which it has not shown up yet but it will oh that's because that's upscales there we go our grids will show up at some point now you can already see here as this is going it's taking more notes from like the colors and stuff of this one but it still centers the person whenever you say like a person standing in the image like on its own without talking about where they're standing or framing or anything it's gonna try to Center them but you can definitely see the influence that our Blended image has on the new describe prompt which is pretty cool so now you can build all of these together with what's called a multi-prompt now I wanted to reference the mid Journey docs because their example for this is hilarious but a multi-prompt basically just tells mid-journey to consider the different parts of the prompts completely separately so like hot dog has two words would generally make a hot dog but hot colon colon to separate the prompts dog creates a dog that would be hot I guess if that's your thing but you see what I mean like these examples are really really good cook cupcake illustration cup cake illustration puts the cake that was Illustrated in the cup cup cake illustration kind of abstracts it a little more so you can do that and you can then add waiting to any of those individual prompts so we can come back in here and build just an insane prompt so we're going to say slash imagine we're going to take our two images that we had as reference images just straight up so I'm going to say copy link image one we'll pull up this one for image two and of course we could apply image waiting to those we can look at our descriptive or our slash describe prompts for this an image of an alien staring at two planets in space in the style of glitch aesthetic this sounds like exactly the kind of thing that we're going for we're also going to add in digital art piece we're not going to do the artist's name or come in here we're gonna press or we're gonna paste this in we're going to do colon colon indicate that this is the first prompt now we're going to come in here get our second prompt in the style of surreal College Landscapes I'm just going to copy this one I know I said it wasn't doing artists but now we give it a second prompt an entirely separate prompt we could also hit colon colon and do a different one so we can do all of these and now we can go in and assign waiting if we wanted to so like I can say a weight of one for this one a weight of two for this one we can do an image we can do image weight of two as well for reference images now we have created the most crazy of Blended prompts here we have our describe prompts from our Blended images we have our original images as references we have a multi-prompt and we have waiting all combined in here we're going to send that crazy prompt out to Mid Journey it's going to show us our two images we put in for reference real quick and see what it spits out this is really layering it on thick of all the different prompting tools you can leverage at your disposal multi-prompting is super powerful but you have to know that each individual prompt gets you part of what you want ahead of time because otherwise you're just throwing a ton of extra descriptions at mid-journey and you don't know what it's going to come up with and again I just want to keep in mind that you can use Affinity Photoshop whatever to build your own reference images or modify the reference images if it's picking up on something that you don't want for example in our referenced little design thingy here if I didn't want that person there if I thought that was you know it describing a person staring at whatever I could go in with clone stamp or object content aware object removal in Photoshop remove that guy just have the grid guy or gal and have it more accurately pinpoint just the aesthetic of the image separate from that person doing all that pre-processing will help a ton all right you can see here we have something very similar to some of the stuff we generated earlier we have lost a lot of that glitch aesthetic which means I want to come in here Maybe copy this prompt maybe put an image weight of one and come over here where it says glitch aesthetic and do an image weight of three leave everything else together see what it says but you get the idea of the process that you have to work through too just continue to make more and more advanced prompts now there's other little additional things that we could showcase real quick so for example you can use versus to make kind of split view images so we can say this first prompt as is but we can start it out with glitch wave versus Desert Wave and especially with like seasons and stuff like that you can get a split view image that mostly has the same composition but has different things on different sides you could do things like we've talked about the no parameter where you do dash dash no and then you can do no text no people no buildings it won't work every time especially the text one especially if you're doing things like magazine covers or movie posters or whatever but it can definitely help okay here's our first modified prompt where we added more weight to the glitch aesthetic and less weight to the image prompt and we definitely get more of that glitch art aesthetic like I love the cowboy look with the glitch Art Space like that's freaking awesome I would probably remove that extra little moon or move it out here or something but that's really cool and then doing the Desert Wave versus glitch wave thing we are getting some really wacky results here which I love yeah it's just smashing those two together which is pretty cool now you you are not limited to text for your prompts by the way you can also use Emoji which could get pretty weird and interesting and it's not always exact how mid Journey interprets it but if we want I don't know say we're very close on this image of an alien staring at two planets in space we can also say looking with a and then put in an emoji with a rainbow overhead make sure there's a space there's a space it will interpret what that emoji or emote or however you want to call it means and apply it to your image here's our rainbow overhead it definitely influenced the style more so than putting a little Rainbow there but some of these are pretty surreal and abstract and I'm gonna probably do something with them because they're so cool and that's pretty cool in general ways just for saving you typing or just experimenting because that is going to interpret it a little differently at certain times but you can also use a microscope so if we say close up photo of a cat or pillar we can also say microscope put in an image of a microscope catcher pillar and often that will end up doing a macro shot because it's treating the microscope as a microscopic view which is a fascinating way to interpret prompts and emoji and prompts and things like that which is really cool here we have our close-up photo of a caterpillar which is already pretty macro admittedly like that's what it's going to assume that you want but then we just did microscope caterpillar and we definitely got some like macro microscopic shots here I realize macro and micro mean different things but you know what I'm saying macro shots photo photography of a caterpillar without having to specify that which is pretty neat you can do all sorts of things with my journey these are just examples there's lots of rabbit holes I'm gonna go down you can also do something that I did not make enough time to talk about here in this course but you can also do logo generation so if you go to explore and just search for logo or logo design mid Journey can generate some logos now a lot of them are going to be nonsense or they're going to break the design rules because it doesn't perfectly maintain symmetry or balance or whatever but if you get something you like you can then use Photoshop or any number of free websites to remove the background and use it for certain things and you can also vectorize it and there's a website called vectorizer.ai where you can try to vectorize an image so let's just say we want to generate really quick rainbow space logo 80s logo design now we'll do 70s logo design so white background we're going to generate a logo here it's a rainbow space logo I have no idea what it's going to give us but we want to make a logo we want to be able to vectorize it so that we can scale it better to use it in other things vectorizing means just turning it into Vector art so the difference between standard Graphics like mid Journeys generating here standard like drawings or photos or whatever and vectors is that vectors are basically just math to tell the computer where how the graphic is made like lines from point A to point B whereas a standard what's called a raster image is just the pixels it's drawn at the size it is and if you try to blow it up or Shrink it down it has to resize those pixels which can result in it getting blurrier or having artifacts or something like that whereas if you can convert something that's already clean enough into a vector then every time you scale it you're going to maintain size better because it knows the math and the colors from point A to point B to recreate the image at each size that you need which is great for working in Adobe Illustrator building letterhead things like that so we got nothing out of space out of this logo but we at least got a rainbow I suppose I really like this fourth one we're going to upscale it now you don't have to remove the background if you're doing the vectorizer so we're gonna pull up in the website we're gonna save this one you could just copy the URL that went in a folder I did not want it to that's fine we're gonna download it over here we're going to go to vectorizer AI we're going to click our image it's going to upload it it's going to try to analyze the image for where you know lines and shapes should go and it should give you a vector file that you can then upscale more cleanly than the original so you see original image vectorized results it may simplify the colors or erase a couple lines a little bit that's not necessarily A Bad Thing given the like error proneness of mid-journey then we can zoom in and you can see here the original image is going to start getting pixelated so if you want to use it for print or for different Motion Graphics for example your original image is going to get blurry and chunky and gross whereas the vectorized version is clean and crisp and amazing and sure some certain gradient points are not going to be perfect so you know you'll want to combine this with still learning art skills on your own but that's incredible that is incredible and for standard raster images you have lots of options for upscaling images I showed briefly there's the topaz apps like gigapixel and photo AI these are very expensive apps from topaz Labs but they spend a lot of time on their upscaling models I'm not going to show too much of them I have dedicated videos on them already and I've shown multiple upscaling workflows on the channel before but these use various AI algorithms to upscale your image and maintain detail preserve faces blah blah that you could use here there's the free app called chainer with a free upscaling algorithm there's also big jpeg is my favorite online AI super res option that's completely free that you can upload your images to and make them bigger too if you need to upscale them for printing I have whole videos dedicated to multiple different ways you can upscale your AI art and clean them up and make them look nice for printing I've printed giant posters with them I've printed standard you know eight and a half by eleven prints and things like that I've even made apparel with my AI art that I've printed on shirts and they look crisp and clear I have a whole playlist dedicated to that Linked In the description and the last little note I want to give you is just to follow other people that are doing cool stuff with AI art artists and techies alike so an artist I recommend following is Josh chap they are an incredible photographer and director already they've been experimenting a ton with AI art but they've also been a photographer for events and for you know really dramatic Moody stuff and their work is just out there and incredible and they've been able to use their work with photography to influence their AI generations and have now been taking mid-journey to incredible levels these are mid-journey images that create incredible styles with and lots of photo realism because he has his own defined style and because he's blending it with his own photography he can just do so much cool stuff with it he retweets a lot but like these are some shots he took at Coachella and he can then use that to generate better AI art as a result and he's been blending those worlds together in a way that is honestly incredible this is something they made with mid-journey too this looks like it's out of a movie it is wild so follow them and then if you're looking for more like prompt hacks and things like that Nick St Pierre does a lot of like 3D rendering stuff but also messes with uh prompting and stuff like that and they have a lot of threads I think yeah they're pinned they're pin tweet has a lot of other threads on how you can explore Styles and gain information from a journey this has been a long course I don't expect you to have watched it all at once take notes take your time maybe watch my video on using chat GPT to build your own mid-journey prompt crafting assistant if you'd like to make your life easier to remember it better and just use the time coach Jump Around play around with it remember just to play around and have fun don't take it super seriously there's a huge world to explore here and I'm gonna have lots of videos here on this analog dreams Channel walking you through it making the tech less scary so that you can get your ideas out as quickly as possible to just make whatever is in your mind's eye I really hope you enjoyed thank you so much for watching and for your support on this channel over the past six or so months now remember to Be Kind Rewind I'm Addie tag me with what you make over on Twitter in my Discord server discord.gg slash people's box or something

Channel: analog_dreams

Views: 59,785

Keywords: midjourney ai, ai art, midjourney, midjourney beginner's guide, midjourney tutorial, midjourney beginners guide 2023, midjourney tricks, midjourney tips, midjourney secrets, how to make ai art, midjourney prompts, midjourney vectorize, midjourney ai art, midjourney logo design, midjourney v5, analog_dreams, analog dreams, midjourney ai tutorial, how to use midjourney ai, midjourney web design, midjourney walkthrough, how to use midjourney, midjourney setup, midjourney guide

Id: 353slEkp2RM

Length: 168min 37sec (10117 seconds)

Published: Thu Apr 27 2023