Create CONSISTENT Characters – Midjourney Character Design

meet Peggy Peggy Palermo best known for her diverse career as an actress in various movies and TV shows you may have seen her in movies and shows such as the Matrix Fight Club Tomb Raider Pirates of the Caribbean NYPD Blue Westworld Apocalypse Now The Lord of the Rings red sparrow and even as an animated character in Dragon Ball Z if this leaves you somewhat confused and you can't quite remember seeing Peggy in any of those movies then that's because I created Peggy in mid-journey that's right Peggy Palermo is a fictional character created entirely in mid-journey without using any external images and today I'm going to show you exactly how I did it alright so before I get started there's a couple of things that I would like to mention first of all creating a consistent character is not easy it is a lengthy process of trial and error and once you think that you actually found the perfect look you will continue to be confronted with slight variations of that particular look that might not be entirely what you wanted to achieve so do not expect this process to work perfectly in every single situation especially not once you start trying to create full scenes I would also like to give a quick shout out to Chris from the all about AI YouTube channel it was originally one of his videos that inspired me to try this approach and actually take it one step further Chris actually has a lot of really great content on all things related to AI so I really recommend you check out his channel if you have time okay time to get started so instead of recreating Peggy Palermo why don't we create an entirely new character right here right now and we're also going to try to make her look slightly different from Peggy because otherwise you know what's the point so let's start off by giving her a name I'm gonna call her Carla and since I really like working with alliterations let's make your last name Caruso I personally think that Carla Caruso has a pretty nice ring to it now you might be wondering why I'm giving her a name at all well I'm not just doing doing this for fun we're doing this for two very specific reasons first of all mid-journey tends to associate a variety of different faces with specific names I suspect this is because it's using a lot of the images from famous people on the internet like actors politicians and people like that of course this doesn't mean that we're actually getting exactly those faces but we're seeing elements of those Blended together and those produce new faces so by using the same name over and over again we're trying to get modernity to show us similar faces if you just say a woman or a beautiful female then you're more likely to get very different faces every time you enter your commands the other reason why we're doing this is because we're going to try to train my journey to associate a specific name in combination with specific facial features as well as a certain look now this does not mean that you can randomly use the name Carla Caruso and just expect to see the exact same image over and over again that is definitely not how this works but if you keep reusing the full description of your character in associate education with that name that it actually does work quite well so in the case of Carla Caruso I've decided that I want to give her bright red hair a braided bun hairstyle so tied up hair in bright red color and she should be wearing a gray leather jacket now there's two ways we could write our prompt you can either write it in an entire sentence like this imagine Carla Caruso is a beautiful woman with braided bun hair and bright red hair and she is wearing a leather jacket in gray color or you could use keywords and partial phrases like this imagine Carla Caruso beautiful woman braided bun hair bright red hair wearing a leather jacket in gray color both these work equally well just pick whichever you prefer or whichever gives you better results personally I'm going to go with the second option simply because I find it easier to read so let's have a look at our first set of images okay great so this is actually already pretty good however at this point of the process we're not just going to take the first image that we like instead I'm going to re-roll this prompt at least two times the reason why I'm doing this is because it kind of determines the path that we're going to take afterwards so it's fairly important that we pick the right image so let me quickly reroll these images two more times I'm going to fast forward this Okay so we've gotten two more sets of images from these three sets of images I'm going to pick the one that I like the most but that's closest to the kind of style and look that I want to have so personally if you have a look at these I'm not a big fan of these circled sort of images I'm also not too happy with these two characters I'm gonna skip that one and these here while they're all sort of okay then they don't have the kind of High hair that I'm looking for so I'm going to pick this one here on the bottom left even though she doesn't have the perfect hair that I actually want I'm going to hope that we get some better variations in the next few images that we generate and but I think the face is kind of what I'm looking for but remember don't get too hung up on all these tiny minute details because in the end as we go along the facial features are still going to change and Vary quite a bit so this is not what she's going to look like at the end this is just basically our initial starting point so now that I've decided on an image I'm going to upscale that image by pressing the relevant button this one is number three so I'm going to hit this button and we'll wait for it to generate okay so once you've upscaled your image you'll be confronted with the usual set of buttons that you always see however this is where it gets really important we're going to start training my journey to understand the type of look that we want and we're going to do that by pressing the feedback button down here with the emoji with the heart-shaped eyes that's the the button we've given feedback and this tells the mid Journey AI we really like this image this basically tells mint Journey hey I really like this image this is a good representation of the prompt that I entered you're on the right track and keep going next we're going to react with the envelope icon and this will tell the mid Journey bot to send us a message with all of the details regarding this particular job so let's quickly check we've gotten the image The Prompt job ID and the seed now make sure you copy the seed because we're going to need this in the next step now go back to the channel and enter the exact same prompt that we entered just before except that you're going to add the seed parameter including the seed value that I just copied here let me quickly do that I'm going to start by imagine the whole prompt from before and all we do is we add a seed at the end and then we go so once the new set of images appears make sure you do not hit the reroll button this is extremely important because if you hit the reroll button it's going to change the underlying seed and we don't want that instead pick the best image from the set even if none of them are perfect in this particular case I'm going to choose the one in the bottom right simply because she has the hairstyle that is closest to what I want so I'm going to hit upscale on the fourth one here it is and once again make sure you hit that feedback button down here it's been registered and then you react to the message with the envelope icon so that we can go to our messages and we retrieve the seed and we copy it and then we go back to our Channel and we start the process all over again so it's effectively rinse and repeat you just keep on going now you might be wondering how often you actually need to do this well technically you should start seeing more consistent faces after six or seven iterations however the more the merrier and in this particular case I'm actually going to do this 15 times and since this process is quite time consuming I do recommend that you use relax mode if you have access to it just to make sure that you can serve your fast GPU hours and optimize between the two now I don't want you to watch me do this 15 more times instead I'm going to fast forward and make sure this is already done and then continue [Music] okay so this is our very last image and I'm going to give it some more feedback and this is what our girl looks like now remember to get the seed from the very last image because we're going to need that as we go forward we're basically going to use that seed that I just copied on every single prompt that we create from now on so over the last 15 images I've been giving my journey more and more information about what I want the character to look like now that I've trained my journey to understand what sort of hair what sort of hair color what sort of facial features clothing and all sorts of other things I want to see it's time to put Carla to good use and put it into different situations this is where things get really interesting but also a little bit challenging as well mainly because there's no cookie cutter way of doing this and it's going to require some experimentation on our end and when you start doing this yourself on your end as well so let's say we wanted to give Carla a matrix like style what I would do is I would use the following prompt as starting point imagine Carla Caruso beautiful woman braided bun bright red hair wearing a leather jacket in gray color in the style of the Matrix and then with our seed this prompt basically tells my journey to put in the style of the Matrix into the concept and while this does change the image it's usually not strong enough but let's have a quick look so as you can see my journey has decided to give Carla a sort of frontal perspective and also give her some clothes that resembles or something that you would expect from the movie The Matrix however it's not quite what we would want we want to have more of a matrix style so what do we do well a more effective way of doing this is to use multi-prompts and weights in which case your prompt would look a little bit like this imagine Carla Caruso beautiful woman braided bun bright red hair wearing a leather jacket in gray color and then we end that segment and give it a waiting of one and then in the style of the Matrix is our second segment and that is also given a weighting of one and then but then we add our seed at the very end this tells me Journey that we want to take the concept of Carla Caruso with all of the details that we described and blend it together with the concept of the Matrix however bear in mind there's absolutely no guarantee that we're actually going to get the exact look that we want so let's have a look what we can see is that we've gotten clearly much stronger elements from what we would expected from The Matrix still the very frontal perspective but the clothes have become a lot more like in The Matrix and also the hair styles are a bit more like a little bit crazier than what we would usually have now usually this already gets us a lot closer to what we actually want however sometimes the weight just isn't enough so let's increase the weight on the Matrix part to 1.5 so it's still the same prompt the only thing that we've changed is that we've switched the weighting of the in the style of the Matrix part to 1.5 rather than to just one the seed Still Remains the Same okay so this one's actually a lot better already however I'm still missing the sunglasses you know the typical sunglasses that you would expect or that you know from the movie you know so let's make one slight change we're going to add the sunglasses to our prompt and now it reads imagine Carla Caruso beautiful woman braided bun bright red hair wearing a leather jacket in Gray colored dark sunglasses and that's how he changed the first segment while the second segment Still Remains the Same with the same weighting of 1.5 all right this one's much better isn't it like this is this is the kind of stuff that you would have expected so you can mold and adapt your prompt and add things if you think that they're still missing or in case the blending of the concepts removed certain things that you think should be there so let's try another one so this time we're not just going to change the style we're also going to place her in a specific location let's put her inside of a jungle and apply the Tomb Raider style to the image so once again imagine Carla Caruso beautiful woman braided bun bright red hair wearing a leather jacket in gray color exploring the jungle and we give that segment a weight of one and in the style of Tomb Raider we give also just a weight of one this time because we don't know how the impact will be so this one worked quite well I think it did change the style of her hair a bit however it's still mostly tied up so I think it's actually still okay okay so now let's try something more radically different and place Carla into a Marvel comic book let me quickly enter the prompt again imagine Carla Caruso and the whole shebang the first segment we're going to keep as per the original and we're only going to add a second segment called in the style of a Marvel comic book okay so this one already has a pretty good comic book look however there's a little trick that allows us to emphasize it even more without having to change any of the weights basically the only thing that you're going to change is we're going to add in the style of a Marvel comic book not only in the second segment but also to the first segment so when my journey proceeds to blend the two segments it's no longer blending two different types of image Styles it's actually blending two images that already have the same style to begin with alright and if you have a closer look you'll realize that this is really really well done it looks exactly how Carla should look in the face as well and it looks absolutely like a like a Marvel comic book the only problem is that Carla no longer has braided bun hairstyle so let's make sure this is emphasized in both segments of the prompt the important thing is that you repeat the same elements from the first segment that you want to see in the image in the second segment but with different words so in this particular case I would enter imagine Carla Caruso beautiful woman braided bun bright red hair wearing a leather jacket in gray color in the style of a Marvel comic book which is then the end of our first segment and then woman with baited bun hair bright red hair wearing a gray jacket in the style of a Marvel comic book and that's the second segment again we add the seed and then what we get is this and voila we have our original Carla back with the right hairstyle that we wanted and also still in the style of a Marvel comic book and if you pay close attention her face looks more or less the same in every single one of these images so this is how you work your way towards a more consistent representation of a specific character in mid-journey however one of the big problems here is that so far we've only been able to create portraits of car Allah and even if we tried adding extra elements to the scene the problem is that her face would probably dominate most of the image there's a way to fix this and it's related to the way we did our very last prompt which was a little bit more complex than usual I will reveal exactly how it's done in part two of this character creation series which I will release in the next couple of days and once it's live you'll find it right here in the corner of this screen we also spoke about text weights today if that concept is completely new to you then make sure you check out this video right here and if you've never used a seed before then you should definitely check out this third video right here as well keep on learning and take care

Channel: Tokenized AI by Christian Heidorn

Views: 141,451

Keywords: midjourney, midjourney v4, midjourney tutorial, midjourney prompts, midjourney consistent character, midjourney consistent style, midjourney character design, midjourney character prompts, midjourney character creation, create character in midjourney, midjourney character art, midjourney text weights, midjourney weights, midjourney multi prompts, consistent character midjourney, midjourney storytelling, graphic novel midjourney, ai character generator, midjourney v5

Id: RXaw79umdv0

Length: 15min 10sec (910 seconds)

Published: Wed Jan 25 2023