Hey everyone, Rodney here at Kleebz Tech.
Today I'm going to show you how to create scenes with different characters and not
have them all mixed up. Normally when you try to generate more than one person in a
scene, the AI will get the details mixed up; you'll have a woman that's bald, a guy with
long hair. But with a few simple techniques, you can get scenes with multiple characters
that make sense. Now in this video, I will be using inpainting and image prompts,
so I do recommend being familiar with those tools and I do have videos that cover those and
many other topics. Okay, so let's get started. In Fooocus, we're going to have this setup.
I'm going on the speed setting, setting my resolution to 1344 by 768. Styles—I don't need
these; we're going to go with comic book and semi-realistic, and V2. Now under the model, I'm
going to change this and use the Cheyenne one, this one here. I've been finding this one to
be a pretty good, interesting model overall. Now in the advanced area, we're
going to check off developer or debug mode and go to the control tab. Then
we want to check off image prompt and inpaint, and then I put this under the control inpaint
tab; we can leave it there. Now at the bottom, you want to check off input image at the
below here, and then in the image prompt area, you want to go down to the bottom and check
off Advanced so we have all the controls. I've already gone ahead and gotten my pose.
I do have a video that covers the poses, my art website, that would be this one here, it's
a couple of videos back. I do recommend checking that out. I'm not going to cover all the stuff
about that, but once you, if you want to use that, you can get your pose, find what you're looking
for, get the angle you want, go to export, set your height and width, and then click on
the export image and get the image that you want. At this point, I have my pose that I
want so we're going to put that in here. I'm going to change this over to cpds because
I find that usually works best for this. At this point, we're pretty much ready to go. Now,
I could go ahead and put a whole prompt in here on what I want, and I am going to do that initially
just to get things going. What happens is, a lot of these things end up getting mixed
up, and so when we go ahead and generate it, we end up with something that we
don't want. This is the issue; you end up with one character different
than the other, different parts of the scene are different because of the colors that
you put in, and it just makes it very difficult. You can keep on generating and trying to
get what you want, but a lot of times, you'll never get what you want, and even with a
lot of inpainting, it can be a lot of work. As we can see here, we have a female superhero in
a green and black outfit that actually worked; it's got a gold star on it. But
now I got all these gold stars in the background. This is not a male villain
in black completely; it's the same outfit, and as so, and as the influence of the green
colors in the background and everything. So, that's where this comes into being a
problem, and what we want to get around. So we're going to stop this because I
don't want any of that stuff. I'm just going to put a female superhero and that's
it, nothing else. It's all the detail I need, kicking a male villain backwards, and
that's all I'm going to put in here, and I'm going to put in a
back alley and nothing else. Now, if I want to describe this, this is where
I'd want to describe the scene more because that's where you can change the background and
everything else because we're going to work on the characters, those aren't as important at
the moment when you're first setting this up, you want to just get that scene in the
background of what you're looking for. So we're going to go ahead, I'm not too picky for
this image, so we're just going to generate that. Okay, so I'm going to stop this because
I'm happy with that one for right now. I don't like these two, but well, I'm
going to see what this one comes out as, and I like this one, so we're going to
go ahead with this one. There we go. So now you say, 'Well, we don't have the
characters we want,' but that's what we're going to work on next. Now that we have the one that
we like, and I'm going to use this one because I don't have all this wind or this fire because I
don't want that in there at the moment. So we're going to go to inpaint or outpaint, I'm going
take this one, we're going to drag that one down. So now we have our base to start with, in order
to get these poses to stay, and I do talk about this in one of my last videos about the text,
and that comes down to this inpaint respective field. What happens is, if we leave this now,
and we were to inpaint this, it won't maintain that pose because it's not going to use that
whole image to do the inpainting. It will crop it down and only use a portion of it, and that's
not going to work for us on how we're doing this. There are ways of doing that, but we're
going to stick with the simple method here. So what we need to change is this
inpaint respective field because we want to tell it, instead of only working with a
portion of the image as it's going to do now, if we were to mask an area, it's only going
to zoom in on that area and work on that, which in most cases is the preferred way of doing
it because you're not trying to maintain a pose. But in this one, we want to maintain that
same pose, that structure from this image. In order to do that, we need to change this
inpaint respective field all the way to one, which tells it to use completely the
whole picture for that reference. Now we can go ahead and inpaint what we
want to replace. So the first one is, we're going to want to replace one character at
a time. You know, you don't want to do too much; that's the key thing when it comes to this. You
work on one section, one description at a time. You do want to overlap these a little bit because
it's not always going to hold the exact same pose. Okay, I can remove some of this, and my spell
check isn't even going to fix that. Okay, so a female superhero in a green and black outfit
kicking—that's really all I need here. But if I want to add more, I could; that's entirely up
to you. The other thing you could do as well is if you already have a design that you want to get
similar to, you can use that in the image prompt. Now, I'm not going to be trying to get this
perfect, just to warn you of that. This is just to show you how it can be done. Now, we can
crank—well, I don't need to crank that all the way up. We're going to keep these on both one. There
we go, so that's telling it we want to influence it all the way through and very heavily with
the weight. Now, it's still not going to get it correct, but it'll get it pretty close,
and then we could go through and inpaint it. I'm not going to do that. I'm not
going to try to get these perfect, but you could do that. Now that we're all set
here, I've changed my inpaint respective field, so let's go ahead and generate and see if
we can change that character on the left. Okay, so we have our scene, our first character.
So let's see which one more closely matches, because that's the one I'm going to go
with. Some, they get the boots right; some, they get the other stuff right. So we're going to
go with this one here. Yeah, that one works fine. Okay, so we have our first
character done. Obviously, if I was picky and I wanted to get those
better, I could work on more inpainting, just certain sections of the image.
At that point, I could go through, and let's say if the boots weren't exactly
how I want, I could try to work on those. Do keep in mind, if you start doing
the inpainting, if you want to keep that pose exact, you do need to use
the inpaint respective field at one, and you do need to keep that in the image
prompt. Now we do want to remove this because we're going to be now doing the next
character, and we don't want to influence that, but we do want to leave that image prompt in
there because we want to keep the pose going. So at this point, let's bring down the new image,
and we will now inpaint this character. Okay, and we have everything else all set, so
we're going to go ahead and generate. Now, we could do the same thing if we wanted to put
something in the image prompt to influence that, but for this one, we're not going to do that. Okay, so now we have our second character into
the scene. I've kept things pretty simple, but you could go however you want with this.
So I'm going to go with this one. So now, I'd say I'd be happy with that. I'll bring that
down to the inpaint area. I'm going to clear that, so we now have our two distinct characters in the
scene, fighting it out. So that's really how you get two distinct characters separately in a scene:
you work on first, I find the scene, and then you can generate. At that point, now you could, in
theory, generate the background completely first without using the pose, and then bring that in and
try to line things up and draw on it, but I find it's best just to start off with that, and then
you can just replace the characters. Give them very generic information, describe the background,
and then you can replace the characters. Okay, so we have our image, but now we
want to have action text. This is kind of boring without it, so let's go ahead and create
some. So we're going to mask off an area here, and we could go ahead and just prompt it
in. One way of doing it, and hopefully, it gets what we want. That's always
a possibility, but I'm going to show you a different way. This is the way that I
did, uh, my last video with text somewhat. So we have our area that we're going to have
masked off, and we're going to leave the inpaint respective field as it was because we do want to
work with this whole image. Then I'll take, uh, go on to any place that does background removal.
I'm going to take the image that I last generated, and we just drop it in, going to remove the
background, then I'll go ahead, and I'll download it, and which I've already done once,
but so we'll go ahead and download that image. Now, then I just go into Adobe Express
or any other image editor. Now, the big thing here is I'm using
the same resolution as the image, that's important. I will then bring
in my image here. Okay, there we go, now it's centered, so now we have our
centered image, and then let's go with 'Pow!' And bring that in, put it at a slight angle, and
then let's see, we're going to change this text. Okay, now we have our text. I can remove these
guys because I don't want them, and then we're going to just go ahead and download, and now, once
we have our image, we go into the image prompt. We're going to drag that in, and for this one, I'm
going to use Pyate Cany. Uh, you could use cpds; use whichever one works. It's hit or miss,
depending on these; sometimes, one works better than the other, so I find it's just use
whichever one, uh, you find works best for this. Uh, we probably don't need it to be a
really high stop at, um, because those, those are bold letters. Now we can go
ahead in here, and I'll add in a prompt, the word 'Pow!' in yellow, comic book
action word. Let's go ahead and generate. And there we go, we have our action words.
Some of these don't come out the greatest, and sometimes it takes a little bit of trial
and error. You can change how the prompt is, as I said, you could try just getting it
without using the image prompt as I did, but I just find that that usually ended up working
better. But if not, you can do it either way. That should help you create more complex
scenes with different characters much better using Fooocus. Now, hopefully, you found this
video helpful and would consider hitting that like button or even buying me a coffee.
I do want to thank the people who have donated since it really does help out a lot.
Thanks for watching, and have fun creating.
