Transcripts
1. DALL E Unleash Your Imagination: Welcome to the world of creative
Possibilities with Daly, open eyes, groundbreaking
image generation software. In this exciting course,
we're going to dive deep into the art of image
creation using AI. Whether you're a seasoned
designer or a complete big, you're in for an
exhilarating ride. Let's start with the basics. We'll take you through
every nook and cranny of the Daly
user interface. Don't worry if you're not tech
savvy, it got you covered. Next up, we'll explore the
magic of prompt techniques. With just a few words, you can unleash the power of Daly to create stunning visuals that will leave your audience in awe. But that's not all
ever wondered how to seamlessly remove parts of an
image and re imagine them. Using Daly will teach you
the art of in painting, where you'll learn
to give your images a fresh AI powered makeover and then there's out painting. Imagine expanding your canvas effortlessly in any
direction with Daly. The sky is the limit will show you how to
do it like a pro. For those of you who create
more advanced tricks, we've got advanced prompting techniques that will
blow your mind, elevate your creative game
to a whole new level. But wait, there's more. Our course wouldn't be complete without a hands on project. You'll get to apply everything you've learned to
create a jaw dropping logo for business using Daly's
cutting edge capability. So are you ready to embark on this exciting
journey with us? Whether you're here to master the basics, unlock
hidden potential, or create something
truly unique, Dali is your creative companion. Welcome to the Dali Revolution. Let's make art together.
2. Dalle Basic Use: Okay, so let's jump
right into day. If you're clueless, just click
Surprise Me and then Daly. We'll put a prompt here and
then you can click Generate. And while the image
is being generated, you will get to see a couple
of tips by opening eye, which you can use to structure or freeze
your prompt better. Now, the more detail
we put in the prompt, the better the result
is going to turn out. As you can see at the top, when I hover on that line, your image is going
to be better when longer specific
descriptions are provided, as shown right here. This is what a bowl of soup that looks like a monster knitted
out of wall looks like. Let's check out this
Ali icon right here. Which will take me
to the home page. And you can browse some
examples here if you want. You can just click
one of the examples. And then that will give you
the prompt for that image. When you get access to
Ali for the first time, you get around 15 credit, which means 15 image
generations or 15 individual prompts that
you can generate images with. After which you have to
purchase credits from Dolly. And the prices keep changing, so the cost of one credit might be different depending on when you are viewing this video. Another thing you
need to know about is the rigid filtering
process that Dolly has. Any generated images
that violate open eyes, policies will not be displayed. Prompts containing vio***t words might not be allowed to run. For example, if there's a
word like fight or a gun. These kind of prompts
would flag your account. If you try a lot of prompts, a query or the prompt is
being flagged by opening eye, and there is a chance that your account might get
banned temporarily. You need to make sure
that you stay away from those explicit or
vio***t prompts. Let's start with a very
simple example description. If you don't know
what a prompt is, let me just clear
that out for you. It's just a text description
that you give to the AI model to
describe your image. Let's start with a
three D render of Microsoft Clippy on a computer
desk running Windows 95. Let's see what's
going to happen. What are the things
that I'm trying to do? I'm trying to tell Dolly that
I wanted three D render, so the graphics should look like they're
three dimensional. I want this to be clip, this clippy character
on a computer desk, and I'm assuming that Dolly
would understand that the computer there is
running Windows 95. This is the result
that I've gotten. I would say this is not a bad result given that
it's our first try. Let's move on from this and then try out something
completely different. Let's say a high
resolution Nasa picture of our galaxy captured
by an in red telescope. Now notice the keywords
such as high resolution an infrared telescope mentioning high resolution won't change the resolution of
the image itself. It'll give us a more
detailed and sharp image. Mentioning Nasa here might give the image a
more realistic feel. Anyways, we want a picture
of our galaxy and we want it to be captured by
an infrared telescope. And honestly, these are
really good pictures. These pictures look very similar to the ones that
are released by Nasa. Every once in a while, you can see the galaxies
are shown in red due to the telescope
being infrared. So I'm going to try
out one final prompt. We get a more holistic
idea how to go about structuring and
phrasing our prompts When using this on line tool. Let's say a minion sitting
on a fast moving yellow cab. Yellow cab in front of
the Golden Gate Bridge. Let's just add the
phrase DSLR photo before minion sitting on a
moving yellow cab in front of a
Golden Gate bridge. All right, the DSLR photo specifies the quality
of the image. The rest of the prompt
describes the photo itself. I want to minion, but I want the minion sitting on
a fast moving cap. So when I say fast moving, the reason I want that
is because there's a specific effect that comes with DSLRs capturing
fast moving objects. And we want the background
to be of Golden Gate Bridge. These are not the
best of images, but you can see it's tried to show that the Gars are
moving at high speed. It's got the background
to be mostly accurate and playing
with the prompt a little might help us zero
in on the image we want. That's it. These are the
basics you need to know to get started with dolly from open AI. See you in the next one.
3. Image Editing: Now I want to show you the
edit feature in dolly. In order to access
the editor function, you click on any image. Click on these three dots
and click Edit Image. This square is called a frame. One of the first
things we can do is if we click on Select
and then drag, and then we can select
and drag this gray frame. If you look here, this
is pan, pan, short cut. You can basically just
click and drag around. And you'll be able to move your work plan to
wherever you see fit, this minus and plus,
it's to zoom in and out. Now that we've got
a hang of that, let's take a look at what
a generation frame does. If I go ahead and click Add Generation Frame
and then I release it, I can actually type
in something new. Let's say adult officers
at an EDM concert. If I click Generate,
you will see that not only did it
generate the image, but we have a chance
to scroll through. And if you scroll
all four of these, you can see that
there's different, various types of this image. What's truly amazing is they're
fused together by dolly. They're similar at
the connection point. Well, let's say we want to
keep this one and we accept. Now let's say we want to add
another frame right here, and this one is going to be a turtle flirting
with a mermaid. We'll click Generate. This is a little bit
troublesome because it looks, it's borrowing from some copyrighted art in
order to create this, since we can see this name here. But if we scroll to the left, we can go ahead and choose
something that's close. This is actually a turtle. If we click the pan
and we move around, we can see that these fusions are coming out quite nicely. Now aside from that, I want to point out this eraser tool. With this tool, what
you can do is you can erase part of an image
and it will fill it in. You can think of it for
Photoshop, for new, the content of Tool, let's say over here I raise the arm and the face and
let's see what happens. And you can see it manipulated this and now it's
done a great job. Although this is an okay, I really like this one though. Let's accept this one and add another frame.
Let's say right here. And I'm going to
click Upload Image. And you can see that I
can resize the image and toggle it back and forth
is a real live image. So let's go ahead with this real life image in addition
to the other three ones. Okay, we've added it, and let's go ahead and add another frame right here
In between all of these, I'm going to type in
a caterpillar dancing in the sky and click Generate. I'm actually pretty
impressed with how flawlessly it's
fusing these together. Now, I'm not giving it any complicated or very
descriptive commands, but it's still doing a good
job of fusing them together. Remember, every single time
you have a generation, you can choose between
a number of variations. So it doesn't just
generate one image, it generates four images. In this instance,
I think this one looks the nicest and it
has a nice color scheme. And also remember that I'm not putting in a lot of direction, I'm only just using
a couple words. With dolly, you can go with
any direction you want. With your prompts, you talk about colors, sizes,
and all that. The last thing that
I want to point out is the feature where you can upload an image and edit
certain features of it. For example, I'm going to use this cat that I have on my PC. As you can see
currently it seems like Dolly only prefers square
images, one to one ratio. And what I want to show is
that you can go ahead and read sparse of your image
and have a prompt here. For instance, I can say a cat with purple eyes
and click Generate. As you can see in
some of these images, it actually did a great job
of turning the eyes purple. In other images,
it just looks like the cat has pink eye and
needs to be taken to the wet. That's how you can
play around with images and dolly see
you in the next one.
4. Use Images As Reference: Okay, so in this video, I'm going to be
taking a deep dive into some of Dolly's features. And let's take a look
at how we can use some reference images so we can import them and use in dolly. Let's look at our
collection section first. So here I'm going to go into
my collections and this is where all of my images
being generated will appear. I haven't any generations
available right now because I haven't generated any images with this account. But for example, let's
say that I asked the computer to me, a photograph, an astronaut
riding a killer whale. And this photo, this is like
something really basic. But what we can do, of course, is to use this image as a reference to actually
create something new. If I edit this image, select that edit image option. I want to edit this
leg because it's on the other side of the
body and it looks weird. What I can do here, I want to make my brush
a little smaller. I want to delete all this stuff that I don't want
to see in my image. I want to replace
that like this. And there we go. Now what I want to do is do
exactly the same thing. I want to type a photograph of an astronaut
riding a killer rail. I want to hit Generate. This feature is
called in painting where we are erasing part of an image and regenerating it. It's going to take the same
pattern, the same colors, and the same image that we just used as a reference to fill
all of what we want to image. Now I'm going to open
the original version, and you can see
here is the r leg. If we open the other
new one, there we go. I think that's a
lot better, right? It's something that's
going to be really useful. We do have more variations. This is another. I do
have this one as well. This one. I think we'll move ahead with this one
because I like the pose. We can do a lot more of this when using it
as a reference. I can go into the
edit section in here, I can choose the option that
says Add generation frame. If I choose a generation frame, extend the image in
any direction I want. If I place it here, I can extend the
image to the right. Let's say for example, here's a photograph of an astronaut riding
a killer whale. What do you want to see in here? Let's say maybe sea creatures
diving through the galaxy. What is Dolly going to do here? It's going to generate an extension image based
on this original image. The more you overlap the frame, the heavier the influence
of the original image. Also, the overlap
part does not change, only the gray part does. Let's hit Generate. This is another way
that we can try to use reference images in order to
add to the original image. Now here for example, we have like an extended
view of the universe. As you can see,
this is something, I don't know what
that is. It's weird. We have this astronaut. I'm going to stick with this
version for some reason. Maybe for some reason
I like this image. Just click, except
I'm going to go back to the history section now. Actually, let's go
to the main section. Click here, I want to upload
my own images, my own image. This will open your files section on your PC and you
can upload your own image. Let's say we upload
this image a bit similar to the one we
generated earlier, more detailed, and
of a higher quality. I'm going to click Crop, and then I can
choose Edit Image. What I want to do here
is I want to edit the image and ask it
to remove this whale. And here I can type sea
creatures and click Generate. This is the same thing we
did earlier where Ali will paint inside the image and it's going to use the
same color palettes. It's going to be like the
most accurate filling tool. I just need to wait
a couple seconds to see what it generates for me. And this is the original right here and this is the new one. I think it's all right. We don't want text in the image really. This one is more cartoonish
and this one is useless. Basically, the phrase C creatures is maybe two
way to get a good output. So we'll have to be more specific with what
we want anyways, if I choose to
upload an image once again and upload
something like this, so I do have this image
here and I want to choose to generate
variations of this, right? Let's click on
Generate Variations. Let's say I want
something related to the original image but
not too similar to it. Here we go. Pretty good. I would say for a first
variation attempt. Here we have the original and then we have four
different versions. As you can see, it's
looking a little different, too different and also
of a low quality. The second option is of course, to edit the original
and painted. Let's click on Edit the Image. And now what we can do is
remove the nose and mouth. Let's say I want to add a young man with no beard, so clean shaven.
Let's keep it simple. It's now generating or recreating perhaps the
person's mouth and face. Notice in the prompt, we never mentioned a nose and
mouth specifically, but Dolly has figured
that out itself. As usual, this is
the original one. This is generated by Dolly. Arguably, this is the
best of them all. But anyways, this
is how we can use reference images to either produce variations
or extend them using frames or just recreate
parts of them. I hope this was helpful. I'll see you in the
next one. Take care.
5. Outpainting & Upscaling: Okay, so Daly is currently
restricted to using dimensions of 1024
by 1024 pixels. And it uses a text prompt
to generate an image. Well, now you can use out
painting to extend the image, so maybe you want to get a
high resolution portrait image of a cyber punk model. Start with a small section of that image and use out
painting to grow the rest. In this case, we start
with the close up of an I. This way the E itself will be a resolution of 1024 by 1024. And everything
beyond that will be added to the image to make the final image around
three to four K resolution. Our initial prompt is a
luxury cyberpunk supermodel, a high definition
cinematic portrait photo. And then we add to
the end a close up on the eyes or even a close up on the left eye
or the right eye. Go through a number of generations until you
get something you like. Then you can come
into the edit section and you can click at
Generation Frame. If you put it in a
totally clear area, it will generate
as usual based on the text prompt to create
a brand new image. But if you overlap it with the current image
as we've discussed, it will look at that and fill in the clear space with
the text prompt. During the process,
you can still use the eraser to
remove areas that aren't working and then
regenerate in the blank space. You can just continue
doing that process to fill out or out paint
your current image. You can do it vertically
or horizontally, and there's a lot of
space you can use. Then when you get a result
that's to your liking, you can now start with
in painting it as this final generation
looks a cyborg. And if you want the face
to be more natural, you can erase the
parts that look a bit disjointed and you can ask Dolly to regenerate them so you get a more seamless and
smooth final result.
6. Face Restoration: Occasionally, when
using dolly to generate images of
people's faces, it produces distorted
results with odd facial features
and misaligned eyes. To address this problem, we can upload the image to the face restoration page of art tencent.com where a
simple process can swiftly rectify these issues
as demonstrated here. As you can see, the face
has been transformed into that of the average Joe
instead of Mr. Frankenstein. As you may have guessed by now, the audio for this tutorial
was generated using AI. What is not in the comments
on how you feel about AI getting disturbingly
accurate at mimicking humans.
7. 1000 IQ Prompting: When certain words are used, images will share
common attributes. One of the most useful
relationships that the AI has learned is the link
between colors and subjects. For example, if you type
in a digital painting of an adorable fire dragon
with no default color, our description,
the machine will automatically make the
default color orange. And if you change it
to a water dragon, then the machine will default to coloring it blue instead. Notice how we never describe the color of
the dragon at all, yet we get completely
different colors by simply changing the subject. Which means if we type
in a coffee dragon, even if we haven't
described the color at all, it's fair to assume
that the, to color it brown because that's the
color of most coffee. And as we've expected,
the color is exactly as anticipated.
But here's the cool part. If I add the words blue
in front of coffee, dragon, watch what happens. So we still have a
dragon and it still has coffee and yes, it's blue. But did you catch anything
else? Look at the face. Do you see that the emotion and the expression of
the dragon has changed? It went from content
and warm to somber, chill, sad, or even depressed. And that's because most
artists in the world know that the color blue is closely
linked with these emotions. The same way that red is
associated with fire, blue is associated with
sadness and being tired. Because for the
vast majority of, for all people on the planet, the sun is yellow, fire
is orange, and red. Plants are green, the
sky and oceans are blue, day is light, and
the night is black. And that has been true for most of our since the very beginning. Now, Dolly, which has been
trained on this data, has learned these connections and can replicate them for you. Which means you can use this to your advantage to
get what you want, even if you don't describe it directly. Here's an example. Let's say you want
to make some posters and tattoo of a
Japanese samurai. For some reason you type in this prompt of a
samurai in a movie, but you want the
poster to be pink. Now most people would just
add the word pink in. And honestly that
would probably work. But another way we could
do this would be to put something in the subject that is associated with the color pink. For example, a princess. Since almost all princesses and princess drawings contain
the color pink, Hence, Ali is associated
princesses with the color pink because
its intelligence is only based on content produced by humans without adding
the word pink at all. If we just change this from
Samurai to Samurai Princess, we should get more pink in
our image just as expected. This is exactly what we get. So when using dolly, just remember that
colors are related to subjects and emotions and you can use them to control
the outcome of your art.
8. Design Lab Special: Logo design with Dolly. Let's get right into it. Let's start with
something like a logo of an Alpaca for a company
that sells cookies. So a logo of an Alpaca for a company that sells
cookies. There we go. Okay, that's not so bad. It creates gibberish text, but it's not bad. You know, it's good
for our first dry. I was expecting less
quality in depth, like this one is reasonable, the text is messed up, but you know it's usable. So let's generate
variations from it. I'm going to do that with this one just to
see if it comes up with something better when I
click on variations and it will generate slightly
different versions of the same image. Okay, this is
actually really good. This one is also pretty nice. Perhaps not like a
super original logo, but it's a cute illustration. Nevertheless, let's try some different prompts
now, For example, a logo for a cookie
company depicting Alpaca. Okay, this one isn't that
good as the other one. Like this one's pretty weird, the face is messed up. I'm going to try to not use the word cookie just
to see what comes up, just a logo of an Alpaca. And I'm going to try being more specific using
some other keywords. For example, a
logo of an Alpaca. Let's add a minimalist
logo of an Alpaca. Let's add the phrase
minimalist to the beginning. As you can see this. Good to be honest Now, this one, this one
is super nice. I tried a new prompt. This one is really nice. I wasn't expecting something
this good to come out of it, but just testing and
trying out different ones, it throws a hipster
way with this one, almost like it's an
Alpaca with glasses. I don't know where the glasses
came from on an Alpaca, but I'm going to run
some variations on this just to give the
AI some direction. And as we can see, these
are looking really good. You look at this one, this is the most cool Alpaca that
I've seen, I've ever seen. So this one is also good, more of a hybrid between
an alpaca and a sheep, but kind of have a
nice feel to it. I'm going to get variations for this one because as I
like this one as well, so it's not that good
in the sense of a logo, but as an illustration, I think it has nice colors. Maybe not yet u refined
enough for a logo. Okay, now we're back on track. So this one looks
spot on, I must say, with its dimensions,
the text font, and the pose of
the Alpaca itself. For some reason, the cookie looks more like a
block of cheese. But overall, we're probably
inching closer to. All right, this one
is even better. This is even better. Notice the dimensions, the pose, the aesthetics, the colors, everything it's honing
in to the image we want. As long as we keep describing
and keep on trying, it's the best contender so far, and running variations on that. We're improving here. We're
improving, definitely. This one's nice. I could
see this going as a logo, so let's try another one, Let's all right. Okay. So look at these
variations are good. They all have that feel
a logo needs overall. This one is looking pretty refined if these
words made any sense, but we can always change
that with minimal editing. It's got nice colors,
nice text fonts, and this one is also adequate, but perhaps not good
enough for a logo. We're going to put that
in the back seat for now, but it's a good
image nevertheless, with its use of colors, the shading, and you know, but it can definitely be simplified or
refined for a logo, so we're definitely on
the right track here. The variations are improving. Like this one looks like an Alpaca had a baby with
a wolf, so quite funny. And this one is also
looking really good, like a dog and an
Alpaca mixed together. So it's making an effort now to write the baker's name as well. Anyways, so this is
how you go about brainstorming ideas
for a logo on Dolly. For now, the text is almost always going
to be nonsensical, but you can get the
illustration by Dolly to be exactly the
way you want it to be. And then you can
just fix the text using a basic editor
like Canva or you can put into the prompt like a logo design of
an Alpaca no text. So trying to tell Dally that you don't have to
include text in the photo. You know, that might not always work, but it kind of helps. So it removes the text and you only get the
illustration and then you can kind of layer the text on top of the
image to get what you want. So a very nice tool for
brainstorming and getting near perfect
illustrations that can be edited slightly to
get a refined logo. So I'm excited to see what you
go ahead and do with this.
9. Course End: Congratulations,
you've reached the end of our Dali journey, and what an incredible
adventure it's been from mastering the
basics of the user interface, to crafting intricate images using advanced
prompting techniques. You've honed your skills and
become a true Dali artist. Remember those
days when removing parts of an image
seemed impossible. Now, you're a pro in
painting, seamlessly, b***ding AI creativity
with your own, and let's not
forget our project. You've taken a
business concept and transformed it into
a stunning logo, proving that with Dali, your imagination
knows no bounds. But this isn't just the
end, it's a new beginning. Armed with the knowledge
and skills you've gained, you're now ready to explore
uncharted creative, or Dali is a tool, but you are the artist. Your unique vision,
combined with the AI, is a force to be
reckoned with as you go forth, continue
experimenting, innovating and pushing the
boundaries of what's possible, keep creating, keep learning, and most importantly,
keep dreaming. The Dali community is
here to support you, and we can't wait to see the incredible art
you'll bring to life. If you're interested
for more though, we have a lot of other
courses lined up for mid journey stable diffusion and lots of other
exciting AI tools. So make sure you keep following this channel for more updates and more courses and more
excitement. Goodbye.
The B., Tech Made Easy!
