Midjourney v5 - Style Prompt Tips and Reference Tricks

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
hey everyone today I wanted to expand a little bit on prompting in mid-journey V5 specifically on posing and image references if you've ever tried to get specific images out of my journey you know that it can be a little hit or miss so through this I'm hoping to give you some tips on how to control your output to get the image that you are looking to get so this video is actually inspired by a comment that was left on my last video by Eric schlitzbeyer I hope I'm pronouncing that right by the way sorry if not Eric Eric wrote what bothers me about mid-journey is that many instructions are simply ignored it ignores sustainably that I want to have a full body picture and not only a portrait from the hip the addition of in the style of is also completely ignored while even Kurosawa and Tarantino should differ at the least I believe we meant Frank Miller should look completely different the result is almost identical for all in the style of editions Eric follows that up with an example prompt a gloomy mystical and already somewhat threatening Forest a 10 year old Viking girl is standing in a clearing where fog is lofting wafting sorry wasting wafting we see her whole body from head to toe she has torn dirty clothes and wears no shoes she threatens with a sword in her hand and screams a loud war cry she looks menacing her long blonde hair is braided in style of and then one of like Tarantino Kurosawa or Miller and while I can't promise a 100 solution to the exact image that he was looking for I think what I find along the way is going to be pretty interesting to you guys so a lot has been made about the new prompting in mid Journey V5 and I think that might be a little overblown the idea is supposedly that it's supposed to be a lot more linguistic and you should be speaking naturally to mid-journey I don't know if it's quite there yet um I still think it's a little more programming and a little less literary but Eric's prompt is great and instantly conjures up an image in my mind so let's dive in and see how close we can get to that so this was my initial output off of Eric's prompt as many of you know without the seed number it's going to come out differently each time this is what I generated these are very cool images but you know we see what the problem here is the prompt did call out for a full body shot we are getting waste up shots I think that usually happens with mid-journey anytime that you're in a 16-9 aspect ratio is that it sort of tends to get very cinematic with its compositions and generally you don't see a lot of like full body shots that's just kind of baked into the natural language of Cinema go watch any movie you're going to tend to find waist up or close-up shots far more often than you're going to find you know full body shots and the other place that this image fails is that at the end of it I wrote In Style by Frank Miller Frank Miller if you're not aware was the comic book illustrator who was responsible for Sin City 300 The Dark Knight Returns a legendary comic book illustrator and writer Miller has a very strong visual style and this is definitely not it in order to rectify that I started moving some stuff around the logic of prompting a mid-journey is that it's going to pay attention more to whatever you have towards the front of the prompt so by shuffling some stuff around in the prompt and applying a prompt formula that I use I covered it in my cinematic mid-journey video you can see that you know linked below don't worry I'm going to go over what the formula is here in a second um we end up with this image which is closer to what we're sort of looking for right it's still not quite in that Frank Miller style but we're sort of dialing in at this point so the formula is a slash imagine cinematic still film by scene subject action set a link or shot and then dash dash but we can use the underlying logic to swap it out to get the results that we're looking for for example in cinematic still I swap that out to illustration film by became Style by Frank Miller the scene subject action and set all remained in Eric's initial prompt the link section which I go into a lot more detail in the Cinematic video doesn't really apply here so I just ignored that and then for shot I changed it over to ultra Long Shot full body and then an aspect ratio of 16 9 and these were our results which I think are a lot closer to what we're looking for particularly that first one um but still lacks a that style of Frank Miller and every one of them has shoes that's one of those things that I don't think that you're ever really going to get over sometimes mid-journey is just going to be insistently ignoring one prompt like I I think it's based on what mid journey is trained on and I guess there aren't that many images of shoeless Vikings so no matter what you do you're just gonna end up with shoes a few more these look really cool for some reason it just gave me a straight photographic one there I'm not exactly sure why but yeah it looks really solid again not Frank millerish it actually kind of looks a little bit like maybe an artist named Becky cloonan who did lumberjanes a very good comic as well um so let's see if we can dial in that frankness a little bit more okay same prompt but after Frank Miller adding in Sin City gets us this which is you know super Sin City uh it almost looks like the overall style of Sin City just came and barreled everything over but yeah it's not what we're looking for here my next attempt was to remove Sin City from The Prompt and instead put a weight to Frank Miller's name you can do that by doing colon colon one which I did for here I think you can go up to five and that just places more emphasis on that particular word in the prompt and we ended up with these images which I think is definitely more in Frank's style particularly I do like that first one it's not still not quite what we're looking for but it definitely now is starting to dial into the look so I did want to try out image prompting a little with this so I grabbed a random photo of of a Viking Warrior uploaded it to mid-journey and then used that URL as an image prompt I believe most of you know how to do that and then added that into the front of the prompt to end up with these images which again are cool but still not exactly what we're looking for the other problem with image prompts is that whatever aspect ratio your reference image is in you then become locked into that aspect ratio but there is a trick around that I'm going to show you that in a second so I'm a big fan of using any tool that's available to you and mixing matching them all together to you know come up with results so I hopped over to Leonardo and uploaded our reference image and ended up getting this out of it which I thought was super super cool now it's a little bit Scurry on the aspect ratio but she's a little squished because of the resolution and I actually ended up uncovering a little glitch in Leonardo where apparently if you upload a reference image that is less than 512 pixels you end up getting weird results like this um just super bizarre and then ended up with another super cool image based off of their image prompt I think they call it prompt magic you can check it out in the video that I did on Leonardo just recently it's yeah it's this is actually a super cool image not what we're looking for but you know still super cool so taking that first image bringing it back into mid-journey uh and then running our prompt yield to just this which again is pretty cool it's still not exactly what we're looking for and that's when I started to think of maybe the problem might be in the Frank millerness of the prompt I removed Frank from The Prompt and we ended up with this which is starting to look a little bit cooler in my opinion I suspect it has something to do with how mid Journey's trained and how it stylizes things for example mid-journey knows samura's very well and mid-journing knows cyberpunks very well so if you tell mid Journey give me a cyberpunk samurai it knows exactly what to do but problems start to occur when you ask for more out there things like a cyberpunk samurai in the style of David Lynch David Lynch has never made a cyberpunk movie or a Samurai movie so mid Journey doesn't really know what to do with that so it just kind of ends up ignoring it Frank Miller has never done a Viking comic I think mid-journey was having a problem applying Frank's style to a Viking comic book so on a lark I decided to try out Mike mingola uh Mike mingolo is the comic book artist who did Hellboy uh and ended up with these results which look actually pretty awesome we still do have the boots on mid-journey is just steadfastly refusing to have Barefoot Vikings but here's something interesting and going back to the idea that style has to have some link to you know the style that you're trying to Output so I rolled back to the idea of putting Akira Kurosawa into this you know Viking girl in the woods and ended up with these images which is not the Viking girl I mean obviously Akira Kurosawa never did a viking film um but look we got bare feet so I did keep trying with that chorus hour prompt just because it was kind of a fun idea I ended up with these images which again are not what we're looking for because now we're in that sort of like close-up range again but they're super cool um and now I really wish that Akira Kurosawa had made like you know a lost feudal Japanese princess in Norway Viking Samurai Extravaganza I would have loved to have seen that movie so just as further experimentation I took that second image out of Leonardo and brought it back over to mid-journey changing out the prompt this time to Ridley Scott I don't know if Ridley Scott has done a viking movie he's definitely done like period piece stuff before but I think his name is a keyword definitely ends up working better than you know Akira Kurosawa and Ridley Scott gives us these images which is actually super cool and really kind of weird if you think about it because we took an actual photo turned it into an illustration and then turned that illustration into a cinematic still inspired by Ridley Scott the future is weird man oh and hey real quick uh if you're enjoying the video please do consider hitting the like And subscribe button I'm very much trying to grow the channel so if you take a second to do that it'll be highly appreciated all right moving on okay so the one big thing I wanted to call out is that Eric's prompt calls out for our Viking girl to be screaming a loud war cry and that's something that's always a little bit tough to get mid-journey to do emoting or acting of any kind um so one trick that you can do is you just do a really bad photo bash for example um I went into Photoshop and I just have this as a stock image and then I pasted it over our Mike mingola output um I mean it's a pretty bad job it's really awful using that as an image reference in the same prompt ended up with this which is a lot closer to I think this one is actually kind of the one that I like the most but we actually have you know some real character emoting going on in the scene so that's something you can try out if you are looking for characters with specific emotions we do still have the problem that our images are in a 2-3 aspect ratio um I think it's two three two three ish and we want a 16 9 image mid-journey is not good at redoing aspect ratios usually end up with images that look like this it'll just kind of like stretch things so again we are headed back over to Leonardo again if you haven't signed up for Leonardo you definitely should they just introduced a paid plan they're still a pretty generous free plan happening right now and they're starting to whitelist users a lot more quickly so in Leonardo we have the canvas feature over here we can simply take our image and drag it in and using this box I can put it like say here-ish or so and then using some parts of our image prompt namely uh Mike mingola and the gloomy mystical threatening Forest we can then start painting out sections of our image to expanded into a 16 9 format or whatever format we want to do it's actually L painting is really cool if you haven't messed with it much you have four options in terms of like what it thinks it wants to continue as like I like that so I just hit accept and then I can move the box over this way and just continue on with the same prompt and then you kind of keep going and just sort of fill out into a 69 image if you have any tips or if you have any questions please let me know in the comments below uh and if you want to see that video on cinematic prompting in mid-journey that's coming up next my name is Tim thanks for watching we'll see you soon foreign
Info
Channel: Theoretically Media
Views: 32,537
Rating: undefined out of 5
Keywords: midjourney 5, midjourney ai tutorial, midjourney ai, midjourney prompts, midjourney prompts tutorial, midjourney prompt tips, Leonardo.ai, leonardo ai tutorial, leonardo ai, leonardo ai canvas, AI Art, ai art, midjourney tutorial
Id: j3LiGOBglCs
Channel Id: undefined
Length: 11min 57sec (717 seconds)
Published: Tue Mar 21 2023
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.