Transcripts
1. Class Trailer : Artificial
intelligence tools are changing the entire
creative space. Adobe Photoshop
has come out with an amazing AI tool that
absolutely blows my mind, and I think it's going
to blow your mind too. This new generative
fill tool gives us insight into the power of
combining AI and creativity. In this class, we'll talk about the new game changing
feature added to Adobe Photoshop's latest
software version. We will get a handle
on the basics of the generative fill tool and
generative expand tools. By doing basic
projects together, we will get a chance to chain somebody's clothes and
hairstyle within seconds, as well as creating a
heart shaped cloud. We'll get a chance to understand
how the tools work by writing effective prompts and creating the right
selections and more. There'll be a few
student projects along the way to try your hand, including a body swap project on a fox and creating a
mythical dinosaur creature. So let's learn about this amazing new
Photoshop AI tool together so we could be on the forefront of emerging
technology advancements and upgrade our creative
workflows in an open Photoshop. My name is Lindsay Marsh and teaching design
theory is my jam. I've been a graphic designer
for over 20 years and a design instructor to over 350,000 graphic design students. I'm excited to be able to
bring this class to you today.
2. Generative Fill Introduction: Welcome back to Adobe Photoshop. I'm excited today to talk about the generative fill option. Now officially part of the
released version of Photoshop, it is no longer in beta. It's now here for all
of us to use and enjoy. I talked about the
contextual toolbar as a new added feature
in Photoshop. So this is just this little
bar that now floats around. It's an extension of
your properties panel. It gives the most popular properties panel
options available, so you don't have to
go all the way to your properties
panel all the time. You have this nice
handy dandy tool bar that you can drag
around and use. I talked about that a little
bit earlier on in the class. I wanted to bring that
up. So if you saw me with lessons not using
the contextual tool bar, that you know that
I was just using a version before this
updated version. So Adobe is always finding ways to make our life easier and increase our
productivity by reducing the amount of time or going
back and forth with options. So with that being said, I
want to talk about it because generative fill is going to be on this contextual tool bar. If you don't see the
contextual toolbar, make sure you go up to
window and make sure that contextual task bar is
checked. Here we go. We're going to go ahead and get started with our first image. I just picked an image of water because I
wanted to generate some sailboats in this
beautiful thing of water. So how do we just pop up
some sailboats using AI? What I'm going to do and shape matters a lot with the
generative fill option, but we'll go over
that in a minute. So I'm going to zoom
in on the water and I'm just going to
grab the lasso tool. We're going to use the lasso
tool a lot with this and I'm going to draw a boat about the size that
I want to have. I'm just going to
do a lasso tool and just do a little boat here. That's why I want
it in the water. I don't want it in the
foreground or the background. I want it in the midground. All of a sudden, your contextual
toolbar will pop up with generative fill and now type whatever you'd
like to generate. This is when prompt writing
really comes into play. And practicing and learning
about prompt writing and other AI programs
can be beneficial. And I have a lot of
experience using mid journey. That has really helped me
learn how to write prompts. But it's a lot easier
than mid journey. I don't have to put on all the specific inputs
like I did there. This is just type a sentence
very plain English. And actually speaking of which, it actually does lots of other languages other than English. So you can type in
Spanish Prompt, and it'll actually work. I think it several languages
that Adobe has used it for. But anyway, so I can
type in very simple, like I'm talking to somebody. So let's say I want a boat and I can do something
simple like a boat. I don't even have to put
a you just I want a boat. So I'm going to press
Enter and it's going to generate three different
options for me. It's generating, and of
course, these little tips they give are actually helpful. So I like to read them because
you have to wait a lot. We'll be waiting a lot
with some of these. And I'm going to speed
up the generation bar a little bit
so you don't have to sit there and wait with
me at the whole time. It generated three
different boats and you can toggle
through these. So that's the first one.
That's the second one which this is going to be the truth
of their generative fill. It's still a work in progress. They're still trying to improve. It's not going to give you
exactly what you want. But of course, all
we said was boat. We didn't say we
want a wooden boat, a fishing boat, a commercial
boat. We just said boat. So it had no clue
what to give us. It gave us three
totally random options. The more specific we
are with our prompts, the more accurate
it will become. Instead of having to redraw
my little selection, it allows you to type in
and refine your prompt. And if you don't
like what you got, you can just press
Generate again and it's going to generate three
additional new images, but also keep your
other three images. And you can continue to
generate and generate until you get 20 options and figure
out which one you like. And you see over
here in variations, you go through all your
variations and they do take up a lot of space
on your hard drive. If you notice you're generating 20 boats and you found
the one you want, you could delete some of these other variations
that you don't want. If you save the
file, your file size is not going to be super large. Now, I have six different boats, some of these are a little bit better and some of them work. Let's say I want to
be more specific. Instead of just a boat, I could type in an entirely
different refined, prompt. I want a sal, wooden snow. Let's see. Let's say I want a
sailboat with blue sails. Okay, let's more specific now. It's only going to
generate sailboats. Remember when I said
that size matters? So when I did this
little selection point, I did like an oval. So now it's going
to do sailboats, but notice how the
sails are put away. A sailboats kind of more of
a triangle shape, isn't it? It's not going to really
generate what I'm looking for. So what I'm going to do is I'm
just going to go ahead and delete this layer and
start over again. And it deletes all
the variations too. Now I want to do is I want
to get more specific. I want to do a triangle
because that's what sailboats saw,
or they're triangles. So I'm going to do a
triangle and it's going to generate within the
triangle but the shape. Matters a whole lot here. I'm going to do sailboat with blue sales and let's see
what it comes up with. It's probably going to generate a better response because I made the shape and what it would be in the end
is a triangle shape. That is much, much better. Let's go ahead and click
on the other ones that ignored my blue sail, but this one, but it didn't
quite get the sail fully up. I could do sailboat
sails fully open, Not sure if that's
going to be better. This is definitely
trial and error. I know some people do
Youtube videos and they only do the
things that work well. And it looks so amazing that it works perfectly all the time. But it's not always going to generate exactly what
you're looking for. It's still a work in progress. Let's generate
some other things. Show the power of this.
Here's what I'm going to do. I want to create a foreground. I want to create a little
landing of grass right here. And you can do any
selection tool when it comes to
generative fill. So I can use lasso tool, Magnetic lasso
tool, anything that selects an object you
can use as your borders. I want to take the
rectangular marquee tool. I would like to have a nice
field of flowers right here. I'm just drawing a rectangle to the generative fill
I'm going to type in. Let me see if I can't pin that because that's
getting a little crazy. So I'm going to pin the bar and I'm going to type in a field. Just field. I'm not describing
what the field looks like, I'm just typing in
field super Basic. What I like to do is
start Basic and then I refine my prompt each
time, each round. Hey, that's pretty awesome. You know what, it did recognize there's water there and it, it interacted with the water. That's what's so great about
this generative fill tools. It takes in effect what's
in the background. So it didn't just pop in a
random field, in a rectangle, it decided to integrate
it within the water by looking like it's
going down into the water. That's the first one. They decided to put a man there. I didn't ask for a man, I
just asked for a field, but they decided to
put a man in there. So not all of these
are going to work. That one actually
works out really well. I like this one, so
I'm going to keep it. But let's say I want to have a little boy fishing
because why not? We're learning this tool, So
let's take the lasso tool. Instead of just doing some
random shape like this, I think a fishing is going to fit really
good into the shape. I'm going to create a shape
of a boy that I want to have. It's not going to generate anything outside of your border, It's not even going to
touch anything outside. It's only going to generate within your frame that you drew. I'm going to do a
little boy fishing. That one looks a
little bit strange, just the perspective
is a little off. You can continue to generate a little boy fishing
with a blue shirt. It also will incorporate
the lighting. You see the lighting source
coming from the upper right. It will also integrate that
within your subject batter. Anytime you do a
generation, Phil, it's going to factor in every
image that is below it. So let's do like a little duck right here. Just a tiny duck. And we're going to
type in, you can continue to do this
over and over. You can basically recreate an entire image if you want to. We can add mountains, a duck. We didn't say if it was
swimming or flying, so we want to say a
duck, duck swimming. And it might come up with more accurate responses
than just duck. You got to tell it, what do you want the
duck to be doing? Let's say, okay, I want to
duck a little bit bigger. Let's do it more
in the foreground. Make it bigger and type
in duck. There we go. That is a little duck. You can figure out, notice how it did and noticed
it was on water. We didn't have to
tell it was on water. It just detected the water in the background and it
made a reflection. So I think that's incredibly
smart, kind of scary. At the same time that it kind
of knows exactly how to do the shadows on the foot of
this boy in the lighting. It's just really kind
of blows your mind. So let's say we want to add some bird flying
birds in the sky. Okay, So now it's going to take images of birds in the sky, so it's probably going to
generate a better result. So there's just
some very basics of degenerative fill
option that you can do. And each time I create a new selection and
do generative fill, it creates a new layer that could be toggled on
and off in visibility. So let's say I created
all that and I don't really like the birds that
generated in the middle. I can delete this layer
or toggle it off, all these on and
off, and it still has the original
image at the bottom. So what I want you to do is I
want you to get some water. And I want you to add
some very basic things using the lasso tool. And start off with the
lasso tool and just create some shapes and notice how the shaped will change your results. And try to create a
little bit more of a basic shape of the subject matter you
want to exist in there. So once you'd add boats, anything you want to
add on the water, you can add a dog swimming, You can add a Doc
if you want to. You can add a field, You can add an island in the middle
of it if you want to. So I would definitely do that as kind of your first
little project, get used to just
getting the lasso tool, creating your very first
generative fill object on water.
3. Changing Clothes, Hair and Styles: I wanted to change
away from nature and I wanted to edit and change
some things on this woman. Let's say I want to
change the shirt. I don't like the
shirt. I want her to wear something
totally different. Well, it's possible. I'm going to take the lasso
tool and shape matters. What I'm going to do is I'm
going to draw a little bit outside of the shape
of this T shirt. I want to generate
something different. And it's not going to generate anything outside of
your parameters, it's not going to even touch
anything that would be outside of the selection
of the lasso tool. I'm just selecting what
I want to be edited. If I don't select the arms, the arms will remain down. If I select the arms, it could change
the arms with it. Let's see what result if I
say I want a blue shirt. You can add different
patterns and textures in your prompt as well. It generated a T shirt, but the arms now look a little
bit too stiff and weird. It decided, since I didn't describe what T shirt,
I said it was blue. I didn't really describe,
I didn't say plain. It decided to go ahead and
be creative on its own. There's a plain T
shirt that's great. Out of the three, usually
one does a great job, one is absolutely silly, and there's one that's
interesting and creative is how I realize that. Let's say, okay, let's
try something different. Let's include her
arms this time. Now we're going to do arms. Let's change it
to a black dress, or let's see, a black
button down shirt. Let's be more specific
with what we want. Let's see what it comes up with. Now, we selected the arms. It's going to include the arms. And the change, the arms will
actually change position, but notice the shape. I had it crimped right here, so that arm didn't really have a chance to move
out and breathe. A black buttoned down shirt,
she's now holding it. I think that's absolutely
kind, a funny mistake. She's holding a black shirt. That's really interesting.
So let's try again. You could see the
trial and error. I'm leaving all these
mistakes in this video because I want you to see
how trial and error this is. I'm going to really change
her and I'm going to give this a nice wide
margin to give it a chance to just have some extra room for her
arms to move around. So I'm going to say a
black cocktail dress and there is the first result. Hey, not bad. It changed
her arms though. Those aren't her original arms. It's using somebody else's arms, but at least it's got the
skin tone matched to. Oh, that looks okay, the hands. Ai. Has a tough time with hands. One thing you could do is, okay, I don't like a particular
part of something. I can go ahead and just
select the hand that I don't, I don't like how the
wrist bends like that. If I go like this and I say a woman's hand and
I just press Enter, let's see what it generates. So you can continue
to stack and modify specific generative fill options on top of each other and
just creates a new layer. There's different
hands that I can use. That one's a little bit
better. Oh, that one's scary. That one's a tiny bit better. But yeah, if you don't
like the way something is, you just draw it. If you don't like
the first three, just keep generating using the same prompt or
modify your prompt. Let's say we want to add,
let's give her a hat. So what I'm going to
do, going to take the last tool, shape matters. If I want to do a sun hat, you want to do a sun hat shape, actually looks
like a pirate hat. Why don't we do that?
Why do a pirate hat? Yeah. A cocktail dress
with a pirate hat. That makes total sense. Oh, look at that. Well, what I find really interesting is how
it does shadows. It added shadows onto the
woman's face pretty accurately. Which is, if you do this
manually, it takes so much work. Of course, you're
getting strange results. I think the best one
was the first one. Then we could delete that and we could do a
different kind of hat. What about a crown? I'm literally just coming
up with this on the fly. I don't want her eyes to change, but I don't mind if
her hair changes. Let's do a crown. I'm just going to be nonspecific with the first one. Okay. That one looks pretty
good right there. Now she has a crown, she
has her cocktail dress. I think she's ready to go. But you can see some basics
of how you can change things. You could change specific
features if you wanted to. So let's say we want
to add side glasses. Let's do like a
sunglass pattern. Let's do goggles, or I
guess sunglasses goggles, more of a goggles shape. Whoa, that is creepy. Once again, it really
struggles with human facial features
and fingers. Those are the two things
it struggles with. You can see it doesn't know if I want swim goggles,
safety goggles. Now I need to be more specific. I want white safety goggles and it's going to probably
get a better result. Once again, the more broad your keywords are
that you're using, the more generic results and crazy results
you're going to get. Let's abandon the goggles
and let's do sunglasses. You could see how
this result changed her eyebrows a little bit and you could see it's not perfect. So you're going to have
to re drawl and re draw to really find
that right shape. That's going to be good for
her face. There you go. She looks like a totally
different woman. She's now the queen of
cool in cocktail dresses.
4. Generative Fill and Expand Tools - Dog Project: I found this empty
warehouse that I wanted to teach you how to add
things and modify things, just like we did with the water. We could also do that
with a room photos. But instead of just
using the lasso tool, I'm going to specifically draw areas using the
polygon lasso tool. Let's say I want to
add a window or a, or anything I want
to have right here. I'm going to draw the
shape of a door in the angle that I think it should be with the polygon lasso tool, let's say I would like a
garage door and there we go. The other options are not
quite as what I wanted, but there's usually
that one that fits. There's a garage door. And look how it integrated it with the texture
and everything. And it even added this lip
right here. It's just awesome. Let's do something else. Let's do the ellipse
tool, elliptical tool. We're going to put like a
big massive window in here. Window that looks out to a tropical ocean.
Well, there you go. Well, that one looks
a little fake, but this one, that one
doesn't look half bad. But let's say, oops, I don't
want the window there. I want to move it. So what I'm going to do is I'm
going to get the move tool. You'll notice it has this
background because that's how it adapts it to the background. It's going to need to
sample that background. Let's say I want to put it here and I just have to regenerate. It won't be the
same exact window, but I'll be able to use
the same prompt and regenerate it a little bit further to the
right. There we go. There's a nice window
over there and I can continue to go on and on. I can even add. Let's go
back and get the lasso tool. And let's just draw a
little figure here. You could say
construction worker Mary is standing with a hammer and kind of integrated
within the background. So you can see how you could use a polygon lasso tool as well
and select specific angles. So how I was able
to add that garage by doing the right angle
and generating it. Who doesn't love golden
retriever puppies? That's what we're going to
do with our next project is I feel like this AI
generation tool, like all the other struggle with humans faces and
fingers and all that, but it does a really
good job with animals. I don't know if it's
because there's just more animal photos
out there to choose from from its library or not. We're going
to give this a try. Let's say I want to, I want to put these dogs in
a little stream. Golden retrievers love water. I'm going to do the same
thing I did before with the grass and I'm going to get the rectangle marquee tool. And I'm going to draw it like they're going to
be in the water. I'm going to put
stream, there you go. They're now sitting
in the water. Let's see what other
options there are. What I find really
interesting, look at that, change their entire pal, structure and put them on rocks. That is absolutely fascinating,
how it new to do that. If you go back to this option, you could see how it
made the fur wet. It changed the little
details that I don't think people are noticing or they are noticing
and they're just going, wow, look, even this parts wet. And they were not
even anywhere near the water before. So you
could take that away. Look at that, it's amazing. There's even no pause right now. There you go. You have all
these really cool variations. You can have them on the rocks. There's pause. It just
came up with that. Just because it had such a huge library of photos to pull from,
that's just amazing. Of course, nothing's
perfect, right? Like there's missing
little claws here. But I think that's incredible. Let's say I want to extend
this photo out even more. I'm not sure which
one I like better. Maybe that one, because I
want to show more water. Well, how do I do
that? I'm going to use the crop generative fill tool. They've added this
with cropping. I'm going to go to my crop
tool here in the tool bar. All I'm going to do, I'm going to go up here to
this option up here. And when it says fill, I'm
just going to make sure. I think this is default
is generative expand. And here's what, I'm just
going to expand this out. Let's say I want more water. That's it. I'm going to press Enter and it's going to
extend that down for me. This is great when
you're working with a photo and you need to extend the background somehow to be able to
fit the photo somewhere. And you could just extend that. I've needed this tool
for so long and I'm so glad it's here. There you go. It just created more water. Let's do it again. Let's say I wanted a little bit
of room on each side. You can do it. Press Enter. It's going to fill
both sides up. Just like that. It went ahead
and filled in the sides. It's like the Content
Aware fill tool, but you could do it with
cropping and it does it a lot quicker than before. Now let's say, let's take
this another step further. Let's add more puppies. I'm going to take
the lasso tool, let's say I want a puppy
back here in the background. I want another puppy. I'm going to be very
specific and say I want a golden retriever puppy. And
I'm going to press Enter. Oh, there it is. And what it did is it noticed that the
photo was a little bit blurry in the background
because this is the focal point of the ****. And it noticed it was a
little blurry back here. So it just went
ahead and took that into account and
made him blurry too, to be able to fit in there. That one's pretty realistic
and I think that one is, the tongue's a little off. I think I might
stick with this one. Let's say I want to put a puppy here in the foreground playing. Now there's a puppy
in the foreground. Once again, this is blurry
too because this is the focal point is in the
mid part of the photo, there's a little
puppy drinking water. Of course, we could
say specifically that we just got the puppy
bottoms on that one. Well, we can specifically say what the golden
retriever is doing. Golden retriever,
taking a drink. Golden retriever doing this, an older golden retriever. Just be more specific of
what you want them to do. Here's another one
I did earlier. Once again, you're not going to get the same exact generation
each time you do this. But you could see how the Pal is wet here when I
added the stream. This is not originally
part of the photo. I just think that's
probably one of the most fascinating
things about this tool still
blowed away by it. But I added some different
dogs in this one. But can you tell which one
is the original dog and which ones were added in if you didn't know that these
two were the original? I just think this is incredible. So really enjoyed this one.
5. Additional Mini Projects: I wanted to show you
really quickly how you can change the background or
setting that subject is in. We have this man running
toward the beach, but I want to put him in an
entirely different situation. What we want to do, we have
our contextual tool bar. I'm going to select the subject, I'm going to go ahead and
inverse the selection. But I wanted to tell you some other modify selection tools that are right here on
your contextual toolbar. You're going to go
ahead and click it. And it's going to be
able to allow you to expand your selection contract, the selection a little
bit and also feather it. I wanted you guys to know
that that was an option. If you wanted to modify
your selections, make them a little bit
bigger or smaller, that option is there. They have this really
very nice handy, inverse selection tool right here in the contextual toolbar. If we don't have to go
up here and inverse, we just click on
this little option right here and it
inverses our selection. Now the man is the only
thing not selected. The whole background
is selected. And now we can generate
wherever he wants to be. So we could do a
field of flowers, we could do a volcano,
whatever you want to do. And there he is.
He's walking, right? They actually put him
right on the path. How cool is that
of that picture? Here is another one where it's a little bit less realistic. You're going to have
a mixed bag here. Just like with everything
else we're going to do with generative fill is trial and error of where
he's going to be. But yeah, they put
him right here on the path. This little area. Can I could have probably selected him a little
bit better on the head. I just did the
automatic selection, but you can see how this works. To quickly change a background, we're going to do something a little bit differently
with this. We're going to change
the sky and add clouds. What we can do is I
can go ahead and get the Lasso tool and I'm
just going to some clouds. Let's do puffy clouds. There's a puffy
cloud, pretty neat. Let's go ahead and
delete this one. I saw someone do this earlier. As they did a heart,
they did generative fill with clouds, Puffy clouds. It generated a
heart shaped cloud, which I thought
was a neat result. There it is, your heart
shaped cloud thought. Just wanted to do some
additional examples. Let's say we want to bicycle
right here on the wall. Let's do a bike. Well you know with this I
might be able to get away with polygon lasso tool and let's say we
want to put a bike, a bicycle right here
against the wall. Oh brilliant. That actually
worked really well. It is amazing. Did the shadows below the petal, the shadows here,
you would not know. If you zoom in, you can tell it's
generated because it's not as crisp and that
looks a little strange. But when you zoom out
and you just look at a picture like this,
it's hard to tell. That's just the first
one. You can continue to generate different ones. This is when you get really
specific with your prompts. So I can say a
bicycle light blue from the Fiftlight blue Vespa. And you can come up with all
these very specific things. Let's keep going. I want another window
or whatever that is. We could do a Roman
style window, whatever that would
look like. Here we go. There's a cool
window right there. You can really see the
power of this tool and action Real quick. I wanted to demonstrate
the generative expand option with
the crop tool. One more time just to
show you the power of it. Here's the crop tool
generative expand. I'm going to make sure
it's on generative expand. So I got the crop
tool, and let's say I really need this
photo to be bigger, or I just want to see a little bit more
of the girl's face. There's so many reasons in
design where we need to expand our background
and make it a little bit bigger so we're
just have it selected. And I'm just going to
generate or press Enter. Wow, isn't that amazing? It filled her in it
pretty realistically too. It did in whatever she
has on her head here. It's a little headdress
and it did it completely. I mean, that's amazing. So
this is before and after. And it decided to
do this intricate, that's all made up down here. That looks like a
really nice that she has on completely made
up by the AI tool. So I just wanted to
demonstrate that again how massively improved this is
over what we had before. We had to do the content aware fill and do
all that stuff. This is just so much better.
6. Body Swap Project - Fox: Now for the next
generative phil project, we're going to be able to
turn this cool photo of a fox into a robot
fox because why not? We have those capabilities. Now I have this photo. If you want to
link to the photo, I'm going to put
it in the document that came at the beginning
of this section. Hopefully it'll still be
available, it's on pexel. If the person takes it down, then you can just
find another photo. But what I like about
is you get to see the whole animal and you get
to see the ground as well. So we're going to get to be able to change a lot about it. Let's go ahead and
unlock this layer, and let's get started. What I want to do is I want
to change the fox body. I'm going to actually get
the magnetic lasso tool, and I want to just
click, and I want to be able to preserve
the fox's face. I just want to do the body, and I'm going to draw all
the way around the body. I'll leave a little room for the generative fill to
have some room to edit. And I'm going to put robot
body and see what it does. Wow, that is definitely special. Every time I do this, it comes up with something
totally different. I'm going to generate
one more time. That's not half bad. Let's
see what else it generated. I like that one the most. Oh, wow. That's almost scary. Wow. That looks more
like a human shape. I think the one I liked.
Oh, I like that one. That's really interesting.
Gave it the dog legs, it detected that it
was some canine, of course is a fox, not a dog, but I was able to detect
that in some and not others. Let's go back to
the one I liked, which looked more like a
superhero dog than anything. I think that's cool with
the metal arms, let's say. I don't really looks a
little messy back here. And it would be neat to add
like a little backpack. I'm going to get the magnetic lasso tool and I'm just going to draw the very top layer and
draw maybe a backpack shape. Let's just say backpack. We can add a little
backpack to him as well. I think we'll stick
with this one. I ended up changing the prompt
a couple times and I said, an old computer backpack. I don't know if that looks
like a computer backpack, but I had to go through
about ten different prompts. Of course, that would take
about 10 minutes to show you, but I ended up coming up
with old computer backpack. The backpack that I wanted now, our little robot dog Fox, I should say, is coming along. The subject matter is
really heavy to the left, and I want it to be
more in the center. We're going to
expand it with the crop generative expand tool. We just have the
crop tool selected, or make sure it's on
generative expand up here. And we're going to expand
this out a little bit to make sure that the fox
stays in the center. As the subject matter
balanced, of course, you also get three options
and you can pick which one is the most realistic,
maybe that one. Now what I want to
do is I want to make this a more dramatic
background, right? We have our superhero Robot Dog. Let's do a little puddle. Let's do the Lasso Tool. And I'm just going to
draw a little puddle on the floor and type in
puddle, nice puddles. Let's see which one we like the best just adds a
dramatic reflection. What I think is very
fascinating is just like the dog's paw and the water, it took the newly
redone robot dog and reflected our fox and reflected
that into the puddle. You see that? It's amazing how it knew to do that,
that it was water, it was reflective
that it would have the sense of shadows
and space and lighting. It takes all that
from the background. Whatever layers are below
it is, places it on. Let's add another foreground element to have the lasso tool. And I'm just going to draw
this hold down shift. I'm just going to draw,
let's just say mountain. It just adds a little
foreground element here. Let's add another one here. We could say Rocky
Cliff Mountain Cave. What if we did cave? All right, so what that does, it just adds foreground
and background elements to help frame and
center the focal point, which is the fox. There we go. Now it looks like he's at
the entrance of a cave. I just find this
so fascinating how it was able to
adapt the lighting. You can put the
dog into a stream. This is just so incredible. Just for fun, I just
did the rectangle tool. Let's see if I could
do the rectangle tool. Let's say we could put him in a stream then it'll
just change everything. I'm just so blown away
by this technology, but I thought this was a
really neat project to do. Now the fox is adapted
into the stream, but I like it without it. But you continue to mess with the backgrod
if you wanted to. You could change anything you want about the body or the legs. You just keep adding
layers upon layers. You can see how many
layers I created just in this project alone. And you could toggle
them all on or off, depending on what you want. So there's our little robot Fox.
7. Head Swap Project - Dinosaur Man: Let's try our hand at
yet another project. This time we're going
to turn this man into a reptile creature with
rainbow colored hair. That's right, We're
going to do something pretty crazy with this one. I'm going to go ahead
and put a link to this, to the Pexels link. So you can download
this photo or find something similar where we
can do a head replacement. Here's what we're going
to do. I want to go ahead and isolate this image. I'm just going to
go ahead and unlock this layer and do a quick select subject
on my contextual bar. And I'm just going to
go down and mask it. I'm going to go ahead
and add a background. I want to do a similar color, so I'm just going to highlight one of these orange,
yellow colors. I'm just going to do a rectangle tool and I'm
just going to make this like a solid color and
put that in the background. Okay, here we are. We're going to go
ahead and do this. What I want to do is
isolate the head. Let's probably
magnetic Glassotolill. Be the easiest to
select our subject. I just want to select the head. I want to keep everything else. If I want to do like
a dinosaur head, let's do a dinosaur head. I want to make sure
I'm giving plenty of room for that dinosaur
head to come out. Because if I don't, it's
going to do this real stumpy. It's not going to
have enough room to really fill that in. Each time I do this is different because each shape
that I create, if I make the shape this
big or cut it here, it's going to generate
totally different results. You have to experiment
with a prompt as well as the shape dinosaur head. It's probably give me
a different result than when I was first
experimenting with it. To practice for the class, we'll see what happens,
that one's creepy. Interesting. And you're
not going to always like every aspect of each head. That's okay, because
we could just draw the parts we
don't like and re edit if you don't like
a particular head. We could just do
this again and maybe do a slightly different
variation of that shape. That one's creepy. Got that
horn coming out of it. Interesting. I ended up drawing a little bit of a St because everything was
getting really long, so I just made a little and
was able to get this one, but I'm not very happy. Like I love how the teeth look. That was probably the best one. After generating, I would say maybe I had 20 different
options and that was the best. Add up the three, here's the other ones that it generated and it
just looked messy. And this one looked
the least messy. Just you're going
to have to do a lot of generating over and over and over to
find the right one. It looks like it would just pop up with this the first time. That's not the case, takes
20 or so different times, but you eventually get to something that
you can work with. I don't like the eyes at all. So what you could do
is just going to draw on top and I'm going
to change the eyes, let's just see
what it generates. That's a little bit better. That one I want to make the eyes maybe even bigger
so I can just create. Go ahead and delete that old
one and see what it does. If I make the area a tiny
bit bigger, that could work. Let's do that was a little different than the
example I had before, but I like how it pops
out a little bit better. Let's draw some rainbow hair. I see this little bit
of hair on top of here. But I don't like it and
I want to make this a really interesting
mythical creature, so I'm going to make sure I
got the top layer selected. I want to do layers on the top. And go ahead and
draw rainbow hair. I might be a little too
much of a selection, but we're going to
figure this out. Rainbow hair, It's a
neat mohawk thing, pony tail. That's
interesting too. Let's go ahead and
generate again. Ooh, that one looks great. I don't like any of those. Let's add something interesting. Let's do this and say
rainbow hair with horns. Let's go ahead and
delete that layer. And we've still got
our rainbow hair. Let's just do horns in the rainbow hair
that we already have. Because I like how the hair is. Let's do horns. And I have to remind myself that shape matters a lot in this. What I need to do is if
I want real spiky horns, I probably need to go
out a little bit here. Let's do rainbow horns.
That's interesting. I finally found
some horns I like. It took several
different generations, but I finally found horns that I thought looked realistic. But I want to create
hair that also goes down over the shoulder
on the other side. I'm just typing in rainbow hair and did a lasso
tool on that side. Oh, look at that. Perfect.
Look how it over laid it over there and adopted the
shadows and how it lays. So awesome. I love
it when it works. It doesn't work a lot, but sometimes when it
works it works. I don't know. The first one and the second one are both good. Let's stick with
that one. I like how it lays over the shoulder
on that one better. I want to extend this hair further and you just
keep going and going. I did one of these earlier and I had probably 30
different layers. You're just build over
on top and you're building layer after layer
on top of each other. So you want to make
sure each time you do a new lasso tool that
you're in the top layer. Because if you do a lasso tool or a selection of any type, and you did it down
here on this layer, it only adapts to
the background. That's the layers below it. And it'll ignore
everything on top, so you want to be on
the most top layer. So it takes all these layers into account when it
does its generation fil. Let's do a, a cool braid that comes down
over his shoulder. And I might need to
make that bigger, but we are going
to trial and error this board Maine bow hair. Well, let's just do braid. I'm afraid if I do hair
it'll ignore the braid. Here is our rainbow hair braid. And boy, it took a ton of generations to finally
get the braid. I ended up just doing
rainbow hair and somehow, but just happened to
do a braid with it. Boy, it took so long, I wanted to leave out all the
times I've tried and tried, and but this is what it
finally came up with. Okay, so we have our braid
or rainbow hair or horns, really looking like a really
interesting creature, But we have these human hands. Well, that's a problem because
that looks really silly. It looks silly anyway, but could look a lot cooler if we get rid of
these human hands. Let's go ahead and
select our main layer. Let's do the magnetic
lasso tool with this one, because I just feel like I
can select better with that, just going to draw
around the hands. But what I want
to do, instead of just tracing the hands exactly, I want to have some room to give the hands a chance to grow
and be a little bit bigger. What I'm going to do
is I'm going to trace. I'm going to make sure I get
all of the hands in there, but with a little extra room. I want to make sure the sleeve
is in the same position. Now I'm going to
put dinosaur hands. I had a couple failed
and funny generations. What I need to do
now is I realized I was selecting using this layer, but I never moved my selection. Here's what happens.
I know this is a little complicated,
so hang in tight. Let's make our selection again. Do a nice generous hand space. I left my selection here. See how I'm in this
layer down here. What it does when you
do generative fill, it takes into account everything
that's below the layer. So it's going to take into
account this picture of the man and the
orange, and that's it. It's not factoring anything
else we've created. It's ignoring all of these new layers on
top that we created. What it did is it ignored the fact that there's this
texture from the face. What it was generating was just random stuff not related to the things
we've already generated. And we want this to be thematic. It'd be very helpful if it knew some of the other things that
was part of the picture. To change this, what we want
to do is we want to go up, we want to select higher out, we want to be higher up
in that layer selection. We're just selecting
the topmost layer, then we can do it. And then it's going to give
us better results because it's going to take all the
layers that were above it, are below it into account when coming up with
the textures it uses. That's what makes it smart. Let's try again. Finally, I got
something I wanted. I tried T rex hands, scaly hands, alligator hands. Finally it ended up just being dinosaur
hand over and over. After 24 samples, I got
something I can work with. You can start to see
the unpredictable nature of this tool. Let's do the other hand, let's go ahead and make our selection. Let's go up top and
do Dinosaur Hand. I was struggling to
get this hand to work, so what I'm going to do is
I am actually going to get the lasso tool and draw a bigger section, a
bigger selection. I'm going to include
the whole arm. It just has a little bit
more space to work with. A, a little bit better. I don't like this random
stuff it generated. But it's better results
more realistic looking. It must be taking a
sample from this braid. What I could do is delete
this and go under the braid. I'm going to go under the
rainbow hair right about here. It's going to take into account all this dinosaur
skin and texture. But it's going to
ignore everything above it. Like the rainbow hair. Maybe it won't take a sample
of that rainbow here. I don't want that in the
arm. So let's try again. Trial and error, I'm
getting better hands. Maybe some of these
look alien like. Maybe that hand,
it's not the best, but I had a little bit better. This is the project
I generated earlier. It looks a little bit different. You're I can't replicate
this exactly how I did it, but this is what
I'm trying to do, replicate for you
guys in the class. But when I did it,
I really liked how this hand went and this
was the selection. This is the whole. I
did the same thing I just did and was able to get
that arm to bend like that. That was the shape I used. Another thing I want to add
is a cool tail like this. Of course, our tail that
we generate will not be exactly like this one because
everything is unique. Let's go back to ours that we're doing here and let's
add a cool tail. Let's go on the top
layer, start there, get the lasso tool and have this really cool
tail that comes out. We could say dinosaur tail. Then we can get more specific if it doesn't generate
what we want. That's like a
little skinny tail. It's a little bit more
skinny than I want it to be. No big deal. Just going to
delete the shape and try with a thicker selection. That's the best tail
I can come up with. I just generated three times, but I got some really
weird tail results. Sometimes you'll just get a dud where it doesn't
even have anything. Whatever it generated
was too small. You will run across
that sometimes, but I figured that's probably the best one
of all the bunch. I still like the one on the original I did
with the zebra pattern. For this one, what did I use? You can go back on old layers and see
what prompt you use. I just use Dinosaur tail. For some reason it generated
a better result for me. Maybe I needed to do
a thicker selection. There you have it. Our first mythical
creature that we created. I want you as a student, to do a student project
to take an animal, or a dinosaur, or some kind of living creature and make
them a mythical creature. You can put rainbow
hair, wings, tails. I don't care what you
do. You could take two animals and
mix them together. I want you to experiment
stacking different generative fills on top of each
other to kind of craft your own creature. I'm excited to see
what you come up with.
8. Wrapping Up and Student Project: As you can see, this is both a work in
progress for Adobe. They're still working on it. It's only going to get better in the next couple of years. But it's also a really good
insight into the future of how AI will affect
graphic designers, illustrators, and
creative people. That it's not necessarily something to be
incredibly afraid of, but maybe something that
might start to excite you. It can change our jobs
dramatically so that we're able to create so much more
things and a lot quicker too. It doesn't mean that our
jobs are less valuable, it just means our
jobs slightly change. And I think we get to be
more creative thinkers with these tools instead of having to get down
into the details. So I'm excited
about the updates. Hopefully you enjoyed
this section. I look forward to seeing a couple of the student projects. I want to see what you
create with this tool. It's unpredictable, but it's
also really fun to use. Don't get frustrated If you
don't get the right result, it takes lots of
time and lots of prompt work and just generating
over and over again. You saw how I even struggled to recreate something I did before because not everything's
going to be exactly the same each
time you do something, and that's what makes it fun.
Lindsay Marsh, Over 500,000 Design Students & Counting!
