Midjourney 6 — The Biggest Update You Should Try!

mid Journey just got updated to version 6 and it might be the biggest update yet every part of it got improved and some tricks that you knew won't work anymore so watch this video to find out what mid Journey V6 can really do the Alpha version of the model became available December 21st to all users the list of improvements may not look very long but trust me it is huge the biggest thing for me is by far the ability to generate text and I will surely test it all out there are two ways to access the updated model the first one is by adding D- V6 to your prompt however there is a catch if you type V6 without a space between the two symbols it will not work the correct way of inputting is by typing d-v space 6 and would you look at that here are the four images of a happy puppy it in a snack I can't say that the images look perfect but at least they are accurate enough in depiction of the dogs the snack in their moue however is either non-existent or all crooked but hey that's mid Journey right there personally I think there's absolutely no reason to do things this way why would you ever need to use the old model if the new one is available so the better way is to type in settings and in the dropdown menu switch the model from the default one to V6 Alpha now let me show you how cool this text generation is and for it I'm going to use the suggested prompt from MJ developers MJ M Journey you get it so imagine a photo of the text hello world written with a marker on a sticky note AR 16x9 in this prom we can already see two important things one all text should be in quotes if you use other quotes than these ones it will not work second AR 16x9 is used to specify the aspect ratio of the generated images now let's look at the images themselves surprisingly the text in all four is correctly spelled you will see in today's video that this is not very common for Mid journey to correctly spell everything on the first try but here with this simple text everything looks okay the style of all images is different and I can't really pinpoint my favorite one H has its own charm be sure to tell me which one you like more in the comments another big Improvement done to Mid journey is more accurate prompt following and accepting longer prompts so let's test that here I have a really long prompt that sets the stage of a bustling medieval Marketplace in the prompt I describe various elements like fruit vendor Storyteller blacksmith details in the background and in the foreground there are a lot of things happening in this prompt but can MJ really deliver and while it is generating let me quickly quote some of the changes taken straight from the press release prompting with V6 is significantly different than V5 you will need to relearn how to prompt this sounds like a big change but as of right now I don't see any any substantial difference previous versions of MJ weren't exactly accurate or convenient to use so I don't really have a lot of expectations towards this newer version mid Journey V6 is much more sensitive to your prompt avoid junk like award-winning photo realistic 4K 8K personally I think this is a good change because before I saw many people using these nonsense words and resolutions hoping to improve the results without realizing that MJ generates all images in roughly the same resolution and the only way to make the images clearer is to upscale them and by the way recently we made a really good video about AI UPS scalers so be sure to check it out so the image is done generating and it does really look like what I have asked for however only one of these images depicts the scene somewhat correctly so I will just click here to upscale the image number two and here I have two options for upscaling subtle and creative let's try them both and while the image gets upscaled let's look at it right off the bat I can see the that the general idea is preserved accurately there is a fruit VOR a Storyteller someone on the ground listening to the story and even someone who resembles blacksmith however there are almost no details making it almost impossible to say what is what now let's jump straight to the upscaled images creative upscaling here I see a really good result it is not perfect but much better than what we had before the faces of some characters look more like faces but some things like this dog are completely unrecognizable but again this is much better and even the blacksmith looks like a proper blacksmith at least if I squint my eyes I could guess that as for the subtle upscaling it looks just as bad as the original image maybe the clarity has improved but in terms of recognizing what is what it is still just as bad so I recommend using creative upscaling and according to developers it gives up to two times higher resolution results before I do more more testing I want to mention a few things about this update that the community has found new prompt length is now over 350 words we can specify colors and other details place things where we want them prompt for more than one subject and talk to MJ like it's Chad GPT right now I cannot confirm nor deny this but I think the next test can shed some light I'm going to ask MJ to generate an image of three best friends of different ethnicities and in specific clothing this sounds like a pretty simple request and it is however once again only one image out of four complies with all the requirements only in one of them the ethnicities are correctly shown and the clothing is somewhat similar to what I was expecting to see so I will take image number three and generate variations of it and now all four Images look really nice it's far from the quality of a real photo but definitely more than you would expect now let me tell you about poo the sponsor of today's video poo is a very cool platform for painless video production hiring an actor feeding him renting Studio who wants all that hassle with poio creating high quality videos becomes as easy as hitting a button for example if you need a marketing video a sign up email a video ad product demo or a video newsletter you can now do it like this first select your digital actor browse through Poo's vast collection of One of a Kind avatars to find the perfect one to represent your message each actor brings their own personality and Charisma to the screen and next choose a voice that you like whether you're SE of friendly professional or authoritative tone you are sure to find the perfect match the library of voices is quite vast and regardless of your needs finding the suitable one is an easy task then simply add and adjust your script in real time to bring your digital actor to life no more retakes or timec consuming editing po makes it super easy to create compelling videos and voiceovers in minutes hasslefree to make the videos even more unique you can add custom backgrounds speech bubbles visual effects music and so on and if you want to edit the video further with po it's super easy to export the video voice over and digital actors directly to your favorite EDI and software PO is a One-Stop solution for creating personalized videos I think this service would be especially useful not only to marketing departments but also to businesses video webinar corporate memos training videos employee and client onboarding all this now takes mere minutes to create so if your employees don't read the newsletter or presentations such a digital actor could make the material more personal and interesting or let's say you want to teach something but you are too shy to show your face with po you can create tutorial and explainer videos online courses presentations and lectures interactive learn materials in a couple of minutes all with personality and character and here is the best for p will be available for 72 hours on the appsumo flash sale so don't miss out on your chance and hit the link below to check the offer out as for the next one I think we should properly test all these new arguments that MJ now supports the initial prompt will be super simple a boy is in a parking lot holding car keys but there are no Cars nearby this promp should test how well MJ V6 can understand negatives and only if m I can truly understand what I'm asking for the images will have no cars in them for the first argument I will use the well-known aspect ratio and set it to 4x3 all four generated images look stylistically very different and only in one of them there are no Cars that's what I don't like about MJ it rarely gets stuff right stylistically the generated images look okay but I feel like Dolly and Adobe Firefly would have done a better job here at least in terms of sticking to what I'm asking for for the second added argument is chaos it is a parameter that requires you to type in the number if you leave it blank MJ will not generate a thing there's no given scale for chaos so I think I'll set mine to five and from what I see this argument basically randomizes the output the style artwork and so on and again only one image fully complies with my initial request the next interesting argument is weird let's not beat around the bush and set it to 100 straight away and the results are well I can't really call them weird but they are surely strange at best now the response is even more chaotic than before the art style on two images is very cartoonish and on the other two pretty realistic however all four Images have cars in them which shows that MJ cannot properly understand the request so if you decided to give it a list of things that shouldn't be in the final image it would be much easier to just not mention them at all and hope for the best and another interesting argument is stylized but it works totally differently from how you expect you cannot just type in this argument and the style afterwards I tried it with anime style no luck this argument is number based and can range from 0o to 1,000 here's what developers say about it lower values of stylize may have better prompt understanding while higher values may have better Aesthetics from my testing it seems like if you don't mention a particular style in the prompt itself and just max out the argument MJ will generate something that either looks realistic or super well drawn for example out of these four Images I like the bottom left one the best yes it doesn't reign with the initial prompt but it looks like there is something that resembles text so out of curiosity I will take this image and upscale it creatively and the result is different the art style is still there but some elements got worse the text is totally illegible now and the hands are super weird another argument to try is style raw here's what devs say if you want something more photographic less opinionated more literal you should probably default to using style raw and at no surprise as soon as I use this parameter I basically get photographs not perfect but super photo realistic for example the boy in the bottom right has three fingers on one hand and only the top right image does everything I asked for there are no Cars in that image but everything I showed you is not the full extent of MJ's capabilities I'm sure you didn't know about the next one do you know those situations when you find a perfect photo of a Panda online and you just cannot live without having the same image as an oil painting no worries because with MJ you can fix that I found out that if I just copy and paste the link to the prompt and ask MJ to turn it into an oil painting the result will be quite good the strangest thing is that MJ doesn't really use the original picture it analyzes it and then draws similar images look at the original photo of a penda and these variations they all have some similarity to the original photo but none of them is a direct copy in another style maybe changing the style would help so I will do the same thing but ask for a d venci sketch style after waiting for a few seconds for the result I once again see that the images are not really all that similar to the original photo and they don't really look like the da Vinci sketch but no worries this is MJ get used to it MJ is not the most accurate out there but the fact that it can analyze images is something new so if you didn't know it was capable of this thing hit the like button I think it's time for us to get back to testing the capabilities of text generation because that's the most interesting part for the next test I will ask MJ to imagine a man in a dystopian City standing in front of a giant neon ad saying subscribe to AI master cyberpunk style this seems like a pretty complicated request with a lot of moving Parts as they say and you know what's strange about these results the first generated photo is a straight up copy of the cyberpunk 2077 poster it's basically identical all the other images are also quite similar to that and don't really show what I asked for and the reason for it is the type of quotes that I used I did it on purpose to demonstrate that only one type of quote Works in MJ so I I will slightly simplify the prompt and change the quotes to the correct ones the new results look much better all images are somewhat nice but image number three in the bottom left corner looks the best so with one click of a button I will ask MJ to create a couple of variations of the same image this took some time but even these variations are still far from perfect this time the top left photo looks best so I will upscale the image creatively which will give us this image I think this one looks great I like the style the amount of detail the colors and the text itself is correct here which again is not MJ's strong side if that's not impressive enough for you now we can create comics with MJ so to do this I asked Chad gbt for detailed prompt which I'm going to paste in TJ this prompt is for a four panel comic page each panel is described in detail and three of them have some sort of text the text blocks are quite long so this will be a proper test for MJ and it's capabilities let's look at those results I can instantly say that these images are far from perfect as you can see some parts of the text are correctly generated but others are crooked and weird completely illegible the artwork itself looks hella nice however MJ seems to struggle with numbers because in every image the number of panels is different and surely isn't four but image number one in my opinion looks the best so I will just generate a couple of variations of that after a few minutes of waiting we get four more images among all these photos the bottom right looks best it has a very consistent style and looks super nice however even after upscaling it the text is still legible so I'd say that creating comics with MJ still requires you to manually type in all the text you want this kind of defeats the whole purpose but will do for a custom Comics book as a present from what I see MJ V6 still needs additional polishing for example it would be nice to not use a special trigger in Discord to create images I know there is a web version of MJ but it takes time for it to get all the features that we can access through Discord but don't you worry MJ V6 is only an alpha now and it will get regular updates further improving functionality that's what the developers say and that's what we saw before so I think in a few months we will be able to see what it can really do but maybe we just did everything wrong and used the incorrect prum structure the developers have released an official guide on wring prompts just one day before the update became public so according to MJ developers the perfect prompt looks like this we start by introducing the style of the image next we need to define the main focus of the image like characteristics of the central subject appearance colors and unique features after that we're supposed to establish the environment or context for the subject including the location time of day weather Etc when that's done we should move on to providing composition details like Viewpoint angle and framing next we need to set the type of lighting and mood after all that we must describe additional details like secondary objects characters interactions or their placement relative to the main subject and if we do it like this we should get perfect results if you want me to test it out hit the button and leave a comment the MJ V6 update is really cool and I cannot wait to see all the incredible things it will be able to do after it gets fully rolled out out I think when that happens it will be perfect moment to compare it to Dolly 3 and Adobe Firefly and decide once and for all which AI image generator is the best we already did that for the older versions of these generators so be sure to check out that video too and see you there [Music] he

Channel: AI Master

Views: 18,068

Keywords: midjourney v6, ai art, midjourney, generative ai, ai, midjourney tutorial, midjourney ai, inteligencia artificial, mid journey, midjourney 6, ai art generator, how to use midjourney, ai generated art, midjourney ai tutorial, midjourney tips, how to use midjourney ai, midjourney tutorial for beginners, midjourney prompts, midjourney alternative, how to use midjourney for free, midjourney face swap, how to use midjourney free, ai master, aimaster

Id: KZWK9xwfwbA

Length: 17min 43sec (1063 seconds)

Published: Tue Dec 26 2023