Transcripts
1. 00 Trailer: The AI wave is already here and it is replacing
millions of jobs. You're here because
you want to create stunning images with
mid journey in seconds. But why is it so hard
to use mid Journey? Did you know artists
and designers are usually better at AI image
creation than most people? And it sucks because
it takes years to master the skill to know
what styles to use, what designs to use, and what prompts to
use. Guess what? I'm gonna give you everything they have so that
you can implement the secrets from the
top mid journey users so that you have the prompts, the strategies, and
the examples to get your images done for
your project in seconds. Why is it so important that you get good at using M Journey? Because unless you are
making tons of money, stock photography can
get really expensive. I was paying for Gies images, shuttle stock, I stock, Adobe stock with
just to name a few, and it cost me
over $750 a month. That is $9,000 a year and over $90,000 over ten years
before using AI. I nearly went broke. And if you are just
using only stock images, it may also lead
to legal issues. In fact, Gide's images, the parent company of
Shuttle Stock and Stock, makes a lot of money, just going after people using their images without permission. They send demand letters
seeking payments of at least $5,000 to
drop legal action. I will also never
forget the time that I used a stock photo for
one of my clients, only to have the client
complain that he ran into another ad that was
using the same image. Imagine my embarrassment. This will never happen to
you when using mid Journey, because the AI images created have never
been seen before. You can incorporate
specific colors and logos, branding elements with
endless creative freedom using styles and themes
ranging from photo realistic to abstract and from modern to classic for your
website, your social media advertising or packaging. And I'll give you the
prompts, the strategies, the examples to not only keep
your quality consistent, but to produce them
very, very quickly. This course is designed for everyone with no prior
knowledge required. So whether you're a
newbie, a beginner, or a season pro, here's
what you can expect. If you're afraid of complex
prompts, don't worry. We'll break it down and
give you everything you need to create amazing
images in seconds, relaxed versus fast
hours, basic prompting. Scaling, variation,
redo, weighted prompts, advanced prompting with
images, styles, themes, artists, perspectives,
parameters, aspect ratios, chaos, seeds, styleze,
quality, and more. Are you ready to not just learn, but master, not just master, but finally get that
image you wanted for your project done and receiving
likes, comments, shares, and sales, then jump
right in and let's start your AI image adventure right
now. I'll see you inside.
2. 01 How to Set up a Midjourney Subscription: If you don't have a mid
journey account yet, simply go to discord, create an account, and
find mid journey on it. Commanding mid Journey
is all done by slashes. Like, when I type Slash, notice all different
types of options appear. So we can type Slash subscribe. Then it will give you this link to your personal
subscription page. When you click that, it
will take you to a place where you can see and change
your subscription plan. I'll discuss the plans just so you have an idea
what's right for you. Keep in mind, these
plans may change, but the overall
idea is the same. If you choose the basic plan, you'll be able to make
about 200 images per month. Not a lot, but a good
way to get started. If you are making a
children's book, for example, on average, you'll be making around 100 pictures
to choose from. Some won't be right, and some
won't be consistent enough, and some will be just fine. So 100 pictures in general, which you can totally do with just a basic plan,
like what I have. But if you need to
create more pictures, then the standard
plan may be for you because it has unlimited
picture generation. Now, you need to know
the difference between fast generations and
relaxed generations. Fast generation is
just as it sounds. The image generated is fast and takes a few seconds
to be completed. Relax generation
is a lot slower. And if there's a
lot of people using mid journey at the same
time you are using it, it will be even slower. Basically, the job will be done when the
system can get to it. For example, if you're using relaxed generation in the middle of the night because
you have insomnia, it wouldn't be too bad. But if you're using
mid journey at noon when everyone and their
dog is creating images, it may take up to 5 minutes for the image to be generated. Which is still faster than hiring someone from Fiver to
do it for you, am I right? And lastly, all of your images will be
generated publicly, where everyone will
be able to see it. So if you want to
be able to create an image privately away
from praying eyes, then you need stealth
image generation, which is only available
in the P plan. The basic plan comes with
200 fast minutes per month. The standard with 15
fast hours per month, and the pro with 30
fast hours per month. 1 hour is roughly 60 images. If you run out and one
more, you can top up, which you can do by clicking
the buy more Fast hours. And you can see the cost if
you want to buy more hours, which I think you would
agree is pretty reasonable. So that is how to subscribe to mid journey and
getting an account. When you go back to the Md
journey page on discord, you can select any room
to create your AI images, such as the Newbie room. Once you are in a room, you can type slash settings
and then hit Inter. A window will come up to
show you your settings. Here, you can turn
things on or off. You can see, I currently have
a stylized medium turn on, fast mode turn on, and high variation mode turn on, which is my recommendation
to start off with. And that's it. Oh,
one more thing. Working on the discord
platform can get noisy, because many people
are creating images. At the same time, you might
hear be p p p, p p, poop, which is the notification sound here and there when
an image is done. When this happens, it might
drive you absolutely crazy. Here's how you can turn off notification sounds
if you want to. The bottom left hand side, you'll see a gear icon. Click on it. Then click on the notification section
on the left hand side. Scroll to the middle, and you'll see the option to disable all notification sounds. Click on it to turn it
on and work in silence. Cool, right? Well,
that's it for now. I'll see you in the next episode where we will start prompting and getting your next
masterpiece, AI art created.
3. 02 Basic AI Prompt Structuring and Formating: Okay, it's time
to start creating AI images for your social media, blogs, ads, e mail, website, and
anywhere you want to use study images to
promote your brand, your business, and
your influence online. To create an image, I
think you know this, but just in case you don't, you pick a mid journey
room like a Newbi's room, and you hit slash imagine, and type out the prompt of what your image
is going to be. So let's start with
a simple prompt. A large burger with a
windy storm behind it. Click Inter, and the image
will be in progress. Just give it a few seconds, and Val, your first masterpiece is done. What do you think? It's pretty good,
right? Especially with such a simple prompt. AI is able to give us a burger with some nice
textures and details. Now, let's say we want a storm behind the burger
to be more pronounced. Let's create the same image again this time
with this prompt. A large burger with a windy storm behind it
along with thunderclouds. Wait for a few seconds. And there we go. Now you
can see the storm with the thunder clouds
are definitely more apparent with
the first photo, looking like an
atomic bomb blast right behind the burger. Not too bad, right? The
burger is tasty looking, but the mood is so dark. I guess it's a
Halloween theme burger. Okay, do you want to level up your AI image generation
scale right now? Sure, you do. Remember,
the format of your prompt has a tremendous impact on
how your image will turn out. Here is a great way to start formatting or structuring
your prompts. Uh, picture, photo,
painting, of topic, background, or other details, in the style of artist
or photographer's name. Descriptors, parameters. Sticking with our burger
fetish, here's another prompt. An oil painting of a burger with a windy storm behind it in
the style of Venton Vang. We can also adding
descriptors such as AK, hyper realistic,
photo realistic, comic, anime style, and so on, which is a bit more advent, and we will cover them
in a different episode. And our burger is here. Looks like something that
Van go will paint, right? Look at the heavy texture
and the large brush strokes, plus the famous stery night
drawing in the background. That is how you can
start to level up your AI image creation
game by taking your topic, the background, and applying
it with different styles. Why don't we do another
artistic style? Here is a prompt, and oil
painting of a burger with a windy storm behind it in
the style of Salvador Dali. What you need to note
is that the further you move away from the
beginning of the prompt, the less influence the word
will have in the final image. So if Sava Do Dali's
style is really, really important to you,
you will put that in the beginning of the prompt
to give it more emphasis. But in our example, the burger in the storm
is still our main focus. And our Dali burger is done. Definitely, some surreal
feelings going on. With the winds carrying
the burger away, or at the bottom
left hand image, you can see another
flying burger is attacking or seducing our
burger on the ground. Just remember, to keep your
most important keywords and directions for your image in the beginning of the
prompt. To review. The first part of your prompt
is the most important. It forms the image
in different ways, according to its form, like an oil painting, a photo, a CGI image, et cetera. The topic should be complete, yet as short as possible. If you're putting a long
and very specific topic, like a man with multiple
arms and each arm is holding onto a cup of coffee with two spoons in each coffee. If that is your topic, AI will be confused, and won't give you
a good output, so keep your topic short. Even though this image turned
out better than I thought. But hey, I asked for two
spoons in each coffee, which I knew AI
will probably miss. The style of an artist or photographer has a tremendous
impact on the image. You can even use more
than one artist, and it can really create
a new, unique style. Like, in the style of a children's book by
artist Name number one, and Artist Name number two. Now, why would you sometimes want to
combine artist styles? If you are ever
worried about creating an image that is too similar
to one type of artist. Like, creating an image in
the style of doctor Sue, it will look very distinct. Creating an image
in the style of Beatrix Potter is very distinct. But if you combine both artist style and add
more artists into the mix, you'll be creating
something that is truly unique and original. And there will be no
way to infringe on any copyright laws
because what you have created never
existed before. This is also a reminder to not be married to one
particular style. If you're using one
style over and over and over again and can't get
the result that you want, guess what? Change styles. Like, if you're using the Disney animation style
over and over again, but don't like how
the image turns out, change it to something similar, like Pixar animation style. And suddenly, you may get an image that's more
to your liking. Don't worry if you cannot spot the difference on the
images that you create, or if all the images
start to look similar, even though you ask for a
specific animation studio. Remember, AI can be
random sometimes, and to get the look
that you want, you may need to mention
a specific artist in a specific animation studio, provided that artist is
famous enough for AI to know. Now, let's move on
to descriptors. Descriptors are extra tweaks. You can add into an
image to make it unique. Sometimes when you use
a few descriptors, things come up very well, such as cartoon
art, Pixar, anime. The image comes out cartoony, childish like and hand drawn,
which is what we wanted. Other times, you may need to use a large number of
descriptors to just let AI pick which to
incorporate into the image like this huge
list of descriptors. Concept art, cartoon style,
well developed concept, consistent style, digital art, vivid colors, very
detailed, colorful in, fantastical, unreal
engine, cinematic, studio lighting, soft lighting, volume metric,
beautiful lighting, global illumination,
and so on and so on. Yeah, I know. It's a huge list. This is a descriptor that I use for a lot of my advanced images, because these are terms used
by advanced CGI animators. Don't worry. I will
give you this list in the downloadable
resources section. But I know what you're
probably thinking. Wait a minute, Jue.
How can you have soft lighting and then have beautiful lighting?
In the same image. Or if you already have
global illumination, how can you have another screen space global
illumination descriptor? In the same image? Well, what I found
out is in this list, even though it's long allows the AI to choose between both. So if you're not married
to one technique or style, you can have AI mix and
match them to come up with all different kinds
of unique looking images? Last, we have parameters. Parameters are the last part of the prompt that controls
things like aspect ratio. The size of the photo,
is it a square, a landscape, or a
wide screen image? Stylization, how much freedom are you giving AI to
create the image? Chaos factor, which is how much it conforms to
the prompt and more. As you create more
and more images, you'll start to write your
prompts in a certain way. You'll start to structure and format your prompts
in a certain way. So as a bonus, I also wanted to include formatting as a variable that you can experiment with. There are two kinds
of prompt formatting, paragraph formatting
and keyword formatting. For example, notice
this first prompt is written in the
format of a paragraph. I'm using words such as n and
the word D in the prompt. Oil painting of a dog
howling at the full moon in the style of fantasy art,
it's paragraph formatting. The same prompt can be formatted
with keywords like this. Notice how in this prompt, the words are way more direct. Now it is formatted
with keywords. So which formatting is better? Well, let's take a look
at the end result. By the way, I chose
this dog howling, because it's like me, howling at all the burgers that we just
created that we can't eat. Anyways, here is the result
from the two prompts. Notice the information
provided is the same. But how you format
the information also creates different
styles of images. Here's a recommendation.
If the image you want to create is simple, then you can write it
in a paragraph format. No problem. But if the image that you're wanting
to create is complex, with a lot of
different variables happening at the same time, then use keywords formatting
because it's more direct and you'll
confuse the AI less. The big takeaway is not
to just copy and paste a large paragraph and expect
AI to know what you want. Very intentional with every word that you use in your prompt. While at times a huge wall of text can produce an
incredible image, it cannot produce
consistent images. Walls of text can get
you a lucky one off, which at times can be
good for experimentation, but not for speed
and consistency. Remember, prompting AI
to create images is not like writing an
encyclopedia or a novel. It is programming, direct and intentional with
specific keywords.
4. 03 Basic controls after image creation: Hello, my friend. How are your
beautiful AI images doing? In today's episode, we are going to talk
about the basic ways to manipulate your AI image
after it has been created. So first up is
upscaling your photos, Enlarging your image and increasing the size
and resolution. A good rule of thumb
to go by is when you see four images that are
not really what you want. There is a chance
that the problem will be fixed by upscaling it, so you can upscale
all four photos by clicking on the U one, U two, U three,
and U four button. And the final output
may be more of what you had in mind when you
input it the prompt. The second is variations, creating variations of
an image that you like, but with minor changes added. This is a good choice to use if the image is almost there, really close, but there is one little thing
wrong with it. Maybe you don't like the parted plant in front of the painting, or you don't like the
positioning of the sun, or you don't like the
shape of the wave. With variations, you can ask
AI to make minor changes to the photo while keeping the overall look
and style intact. Then we have the redo button. It's there for if AI
completely drops the ball, and none of the four images that it created is right
or even close. This could be if your prompt
is wrong, but other times, it may be because of
poor random generation. Then you just hit
this redo button to generate the
four images again. And don't be afraid to redo many times to get what you want. So let's take a closer look at these images and the prompt
that was used to create it. We have an oil painting, which is what I was
trying to make. The topic, a stunningly
beautiful sunset. The background, waves crashing
on the beach and sand, style, in the style of mbrant, descriptors, rough brush
strokes, abstract, and the parameters, which are advanced variables that we will talk about in
another episode. Following this prompt structure, we can easily change
these images from a oil painting to a photo. So another prompt can be a
perfect photo of the topic, a stunningly beautiful
sunset, the background, waves crashing on
the Bach and send, style in the style of
natural photography, descriptors, AK,
portrait photography. So when we did that, we get photos of the sunset
and beach waves, and we can keep on doing this. Another prompt can be
an animate comic image of a stunningly
beautiful sunset, waves crashing on the Bach and send in the style
of cartoon art. Aime Pixar. And we had the same
sunset and beach waves just a completely
different style. Isn't that cool? We can
continue experimenting. So instead of a perfect
photo of this sunset beach, we can change the prompt to a photo realistic image of a stunningly beautiful
sunset and blah, blah, blah. And notice one of them
did come out right. Although the painting of
the sunset wave in front of an actual sunset
wave is cool and surreal and a little funny at the same time.
What do you think? AI is random sometimes and created this photography
canvas on top of a bed, which isn't bad either. Hopefully, inspiring you to
create other image ideas. To show you how random
AI can be sometimes, instead of an oil painting, instead of a perfect photo, instead of a photo
realistic image. How about just a regular
image of this sunset wave? What would AI come up with? It came up with the
sunset wave in a TV. We can also change the parameters at the
end of the prompt, changing the aspect ratio, for example, and notice that now all four images will be in
portrait photography size. Then you can also do y screen, and all the four images will be reshaped to the size of a movie. Perfectly formatted to fit into that weird TV
that we just saw. Then, we can do a surreal
painting. Here's the prompt. In the style of surrealism, focus on texture and form, oil paint on Canvas with
thick s brush strokes and expressive mark making to build up layers of
color and texture. Color palette achieved through
scrambling techniques. And we get this. Amazing, right? Especially this
bottom right image, making it look like
the artist put a heavy amount of oil paint on there just to get that
look that you wanted. Don't forget to experiment
with upscaling. What is upscaling again? It is blowing up the image to increase its size
and resolution. So let's upscale one of our oil paintings we
made earlier. And wow. Just look at that.
Look at how much better it is from the
original when we blew it up. And this one, too,
it really brings out the color and the
oil painting style from the original image, right? Here's an upscale of an
animate image we did earlier. Looks like it belongs
in an epic comic. Here is an upscale of the
photo realistic image. Notice how the upscaling
really brought out the details in the
waves and the texture. You can almost taste the
salty water, can't you? Here's another upscale of
a photo realistic image. You can see that AI really tried to have a depth of field, where parts of it is blurry, helping your eyes to focus on the places where it is
sharp, the actual wave. Now, this is one of my
favorite paintings. Do you remember how we created a surreal painting
with the heavy oils? Well, when you upscale
that, you'll get this. Don't you just love the texture? It looks so three D, as if the artist spent a
ton of time piling on paint after paint to give it a rough and almost touchable
brush strokes look. And this was all done
in second with AI. Now, let's talk a little about other adjustments you can
do to an up skilled image. Here's an example of an upskill
image of a man receiving strength from heaven with a large eagle in the background. Notice, we have further
upskill options, such as upskill subtle, upskill creative, very
subtle, very strong, very region, Zoom out two X, Zoom out 1.5 X, and so on. And remember, these options will change from time to time, depending on how fast
mid journey updates. Subtle is slight changes. Creative and strong
are minor changes. Like, if you ask for
strong variations, notice how the eagle and
the man's poses changes. But the overall image is
pretty much the same. And if you upscale
the image subtly, you can barely
notice the changes. But if you upscale
the image creatively, the sparks that you see
coming from the man are smoother and
finer. Notice that? Making fine variations is particularly useful if you
are creating images for a book where each
image may be a little different to show development of the story or dynamic motion. Well, there's no real way to predict exactly what
will be changed. So my recommendation is to upscale all the images
that you really like and review over them and
keep on refining it until you see something that matches what you
are looking for.
5. 04 More Style change examples of the same image: Finding your style, and
the way you want to present your vision is
an ongoing process. Let me give you some more
examples of style. Shall we? The cool part is that
in these examples, we will keep the topic, the background the same. We are going to
change the style, and you will soon see
how important style is in telling your visual story. Our prompt for today is a
girl and her dog walking down a country road with farmland and crops
in the background, cartoon art, Pixar Anime, which gives us an
image like this. You can see, they
came out cartoony. Now, let's use a
different prompt of the same topic
and background. This time, our prompt
is going to be long, really long. No. I mean it. L, like,
really, really, really, really long, just
to see what happens. A girl and her dog walking down a country road with farm land and crops in the background. Concept art, cartoon style,
well developed concept, consistent style, digital art, vivid colors, very
detailed, color for is in, high quality,
digital art, Mazaki, award winning Adobe Photoshop,
CMYK color, playful, fantastical, unreal
engine, cinematic, studio lighting, soft lighting, volume metric lighting,
beautiful lighting, and so on and so on. Okay, you get the idea. It's a really long prompt, a pretty much the same thing. A girl and her dog walking
down a country road, except with a lot of
details about turning it into a Japanese TV
show, an anime series. And here it is. Notice how the image is still cartooning,
like the first one, but they look a bit different, really focusing on the genre of anime and specific artist styles and specific technique styles. When you compare the two
cartoons side by side, at first, they two
are very similar. They are both cartoons. But when you look
at them closely, one is more of a
general cartoon, and the other is a
very specific type of Japanese anime cartoon using Miyazaki Award winning
style with blah, blah, blah art techniques. Like the difference?
Let's keep going. Let's go back to a very
simple prompt, girl, dog, farmland, cartoon
art, Pixar, anime. And you get this. Makes sense it's another cartoon, right? But let's compare the prompt of this picture
with the last one. Ah, interesting, right? When your prompt is simple, more emphasis is
placed on the topic. The girl and the dog
are more prominent. But the huge pump
that we just did earlier got us a
fuller image with a lot more things going on in the background with the girl
and dog lost in the image. Then we can try in our
style by Louis Roo, who does a lot of gothic and gloomy type images like these. Using the same girl and her dog walking down a
country road prompt, except in the style of
Louis Ro. We get this. Definitely has that
dark and gloomy style. What about Andy Warhol? Have you heard of his style? They look like this, vibrant and loud colors
screaming at you. If we apply the same
prompt to this style, you'll get this definitely consistent with this art style, loud and colorful and
almost uncomfortable. Now, everyone has heard
of Salvator Dali, right? The melting clock guy,
super surrealism. Applying his style, we get this. Definitely weird and surreal
with some of the images where the girl and the dog are walking in
different directions. Probably what Dali
would have done if you asked him to paint
our girl and dog. Then we can combine
different artists together such as Boris Vallejo, Louis Roo, Francis
Franzta, and Larry Altmre. Put all those artists
in one bucket, and you'll get an
image that has a bit of each of their styles used. Like it. We can also
experiment with other styles. Here's one with just a
regular photography style. The girl and dog looking pretty
realistic, glamour shot, giving the girl more
lighting, brighter, and softer lighting, almost trying to turn the girl
into a supermodel. Cinematic. We have a
movie style lighting, telling the story and
mood of the adventure. Dramatic, darker
shadows, reflections, definitely trying to
paint a picture that something interesting is
going to happen soon. So there you have it.
Did you have fun? Experiment with different
styles and upscale them to really see and
absorb the details. It will help you decide
or at least inspire you to create styles that you may not have
thought about before.
6. 05 Advanced Image Control Weighted Prompts: Welcome to today's episode where we will help
you give finer and finer control on your
beautiful images now that you know
how to create them. Today, we're going to talk
about weighted prompts. I'm going to share with you
the theory and the concepts and then give you examples to show you what it will
look like, okay? Remember, the longer the prompt, the less important
each word means to AR, and the more important
the beginning is. Did you know you can give different parts of a prompt
different amounts of weight? This is done using
double colons. Before I get ahead of myself, do you know what is
the difference between soft breaks versus hard breaks? Comas give soft breaks, separating things with the normal beginning
to end weighting. Meaning the first words
in the prompt has the greatest influence
over the final image. Instead of using commas, you can also use double colons. These are what's
called hard breaks, making what's on either
side equally as important. For example, a cat double colon
and a dog double colon in the kitchen tells the AI to make sure each is equally
important in the image. A cat double colon two and a dog double colon
one in the kitchen tells the AI to make
sure that the cat is more prominent than
the dog in the image. The double colons do
need to be touching the word with no
spaces in between. As you start to add
weight and numbers to give your topics
more prominence. It's important to know that the weights are
relative to each other. It's all done by ratios. 110 are the same
value as ten and one. That's how AI distinguishes
between number weights. It is a relationship of ratios. So there is no need
to go crazy and put a 1,000 weight of a topic if you're not getting
the image you want, because the number is going
to overpower everything else. Remember, it's all relative
to each other as a ratio. You can also have negative
weights to tell AI, you don't want something
in your image. Like in this example, double colon negative 0.5
will indicate the same thing as d d no and the item that
you don't want in your image. As my team and I experimented with different
prompts and weights. We notice that some
of the best images also use weighting granularity, like using weights of 1.2, 1.3, 1.75, and so on. I see a lot of expert mid journey image creators
making small, tiny changes to get
a certain result. Lastly, we can also use brackets
to organize our prompts. Like this example, a dog
walking down the street, double colon two,
a cat in a window, double colon 1.5, P double
colon negative 0.5, where the dog in the
street is a little more important than
the cat in a window. And no people should be
present in the image. Okay, okay, I know you must be dying to see
some examples, right? Let's take a look at some,
shall we? Here is the prompt. Universe, double colon one inside a crystal ball
double colon ten. So we are basically telling AI that inside a crystal ball is ten times as important as the
universe. And we get this. Notice how in every
image, the universe, the world is inside the
crystal ball every time. Yes, one of them came
out really weird, a hand with seven fingers, and we probably
wouldn't use that one. You know, sometimes AI will give you random and crazy
images like that. But what happens if
you switch the weight? Let's use an extreme
number example to show you what's possible. Universe, double colon 100 inside a crystal ball
double colon one. And the image you can see, the universe, the world, really takes up everything. And in some images, we don't even see a crystal
ball anymore because the 100 weight value
of the universe wiped out the crystal balls
importance in the image. L et's dial the universe
down a little bit and use the prompt universe double colon ten inside a crystal
ball double colon one. Now, you can see more
of the crystal ball. The universe is still ten
times as more important, but there's still enough room for the crystal ball to show up. Do you remember we can
have a negative weight? You can use colon colon with a negative number
to remove things. Or at the end of the prompt, you can use a parameter
by adding a space. D, d, no extra fingers. For example, no extra fingers, no extra hands, no
extra text, et cetera? Let's dive deeper and
see some examples of diminishing or removing
certain colors in your images. If your prompt is just Rainbow, you get an image like this with lots of bright
colors in the rainbow. But what if we change
the prompt two, rainbow double column one, blue double column negative one? Notice how our result is that the blue shows up because
it's part of the rainbow, but it is so much thinner. How about if we change
the prompt to Rainbow, double colon one, red, double colon negative one. We now have an image that
has a much muted red color. Red still shows up
because it's hard for the AI to imagine a
rainbow without red, but it's doing its best to remove that red per your prompt. And if we upscale one of the images like I showed
you in the last episode, you'll see the result
is more what we asked for with the color
that should be red muted to a pink color or salmon color if
you like seafood. And this up scaled image that is not supposed
to have blue. You can see the red is
thick, orange, yellow, and green, are all thick, but blue is very thin. To help you see the
impact of our prompt, we can compare it to
the upscale version of a normal rainbow, and you can see the colors in the normal rainbow is
much fuller, right? So when should we use waiting? When you are creating
a simple image, you don't need waiting. But if you're creating
a children's book or a comic book where the
images are more involved, you can use waiting to make finer adjustments to your photo.
7. 06 Advanced Image Control Style and Descriptors: L et's talk about using styles and descriptors
in your prompts. Remember to return to our
original prompt structure, picture, photo,
painting, et cetera, of topic, background, or
other details in the style of artist or photographer,
descriptors, parameters. This episode, we are going
to focus on the last part, which is in the style of
an artist or photographer. And then the descriptors. This can completely change a picture style in all
different types of ways. At times, with words that
might not make sense together, but do tell the AI important
things about what you want. Check out this prompt. A green dragon in a
forrest in the style of Luis Roo, intricate,
hyper detailed. The word intricate and hyper detailed are the descriptor
instructions of the prompt. We get this, which I
think you would agree, has a lot of micro details, and this isn't even the
upscale version yet. Cool, right. Let's
try the same topic, same background, in
a different style. A green dragon in a forrest
in the style of mbrant, oil brush strokes,
neo classical. The word oil brush
strokes and neo classical are the descriptor
instructions of this prompt. And you can see these
are different from the other images with
more brush strokes. Now let's try photography. A green dragon in a
forest in the style of Anto atoms,
cinematic photography. Now, the dragon looks more
three D because that's the style of Anto atoms,
specific photography style. We can start upscaling the
images to really notice a style change and how descriptors impacted
the final image. This is from the intricate
hyper detailed descriptor. The details are mesmerizing, probably would have taken a
real artist at least hours, if not days to draw this image. This is from the
oil brush strokes, neoclassical descriptor. Would you agree, these really do have a painting
style to it, right? And finally, this is from the Ansel Adams cinematic
photography descriptor. These dragons look like
forrest sculptures, crisp looking lines that are usually found in
that type of style.
8. 07 Advanced Image Control Parameters: So, how do you fine tune your images after you
have decided on a topic, a background, a style,
and a descriptor. It is all about using parameters to get that last
image quality boost, and that's why it is located at the very end of the prompt. Parameters are like
programming ish codes, we put acifixes to our prompt to more finely
control how the image appear. To use them, you use two dashes, the parameters, a
space, and a value. Remember, no commas
between variables. Comas are soft
breaks for prompts and are not used at the
end with parameters. The most important parameter that you're going to probably use all the time is the
aspect ratio parameter. The one thing that changes the entire dimension
of your image. If you leave it blank, the default image
will be a square. But you can also change it
to AR space 16 by nine, will give you a Y screen image. AR, space two by three will give you an
iPhone style image. Another parameter that you can experiment with is
the chaos parameter, and it is done with C. It will control how closely
the AI sticks to your prompt, how much freedom you
are giving AI to take your prompt and not follow
it exactly word for word. We do this when we are
experimenting with images and not quite committed on exactly
the look we are going for. You kind of want AI to surprise
you to what's possible. The default for the
chaos parameter is zero, and the range is 0-100. D C 25 will let the AI change
the prompt a small amount. C 50 will let the AI change
the prompt a larger amount. If you select the
value larger than 50, it usually doesn't
have that much of a change except
in certain cases. Another parameter is
a stylized parameter, and this is done by S. This parameter controls how
much noise is generated. Or more simply, how
much the AI allows stylized and surreal changes
to affect the image. The default for the
stylized parameter is 100. The range is 0-1 thousand. D S zero will keep the
image very realistic, closer to the style and
keywords you are using. This is what you will
put in if you want your photos to be as
realistic as possible. D S 100 allows a small
amount of creativity, which is the default setting. And if that's what you want, you don't need to type anything, and it will be
using this setting. D 250 is a medium
amount of creativity. And S 750 is a very high
amount of creativity. Good for surreal or
impressionistic images. Let's talk about some
bonus parameters that you probably
won't use often, but it's nice to know
that they are there. The version parameter is done by using V and diversion number. This is if you wanted
to use an older version of mi journey like version
five or four or three. But if you want to use
the most recent version, just don't put anything, and it is automatically the default. W to spice up your images
and make it look fun. Then try the Ninji parameter. It's done by using
da, da, Ninji. And it will take whatever image you're creating and
give it an anime look. Let's say you have a
prompt that gives you an elegant looking woman in a sleek black dress
sitting at a poker table. Wonder how you can turn this
realistic image and turn it into a cartoon or
anime. Wonder no more. Use the Ninji parameter to instantly turn that girl
into an anime character. In one image, she looks
more like a villain, a boss that you have to
defeat if you are James Bond. And lastly, we have
the quality parameter. It is done using the space
and the quality value. Most of the time, you will use, which is standard quality. Sometimes you can
experiment and try four or five because mid journey is always
updating and changing. This may or may not work, but the Q two is usually always available
as the standard.
9. 08 Consistent images with Seeding Parameter: I can't wait to share with
you one of my top secrets, which is using the
seed parameter. Now, why are we using
this parameter? It is because we are creating some consistent images today. These are the images that you create with a
character looks very similar as to make your audience feel that
it is the same person, a great storytelling strategy. It is done with d, d, seed, and then
the seed number. This seed parameter helps you clone an image from
its original source so that you are likely to get a more consistent image just
like from the source image. You can find the C
number by hovering over the Forum post
top right corner. Click, Add reaction, and
find the envelope emoji, which will give
you the C number. I know. It sounds more
complicated than it really is. So let's do a demonstration
together, okay? Let's say you accidentally
created the love of your life, and she shows up like this. Ah. Beautiful and gorgeous. Right? So, how do you clone her and make
more images like her? We can go to the top
right hand corner, and we'll see a button
that says, Add reaction. Click on it, and you can
type in the word envelope. Pick the first envelope
emoji that shows up. Now you can see we have
sent ourselves a letter. Now, we can go into our inbox, go into the re tab
and find the image. It will tell you the prompt
that was used to created, the job ID, and most
importantly the SD number. So now we can go back
into May Journey, use the same prompt as before, so we are more likely to
get the same type of image. The prompt was elegant woman
in a sleek black dress, sitting at a poker table surrounded by golden
chips and cards. And at the end of the prompt, write d seed space, and the seed number. And then you can wait for the AI to try and clone the
love of your life, and you cross your fingers and pray that AI won't mess up. And in a few seconds, you see v. The girl on the lower right hand corner looks very much like her, right? Let's upscale and do a
comparison. Oh, wow. Wouldn't you agree? It really do look like it's almost
the same person. The hair is a little different. The eye is a little lighter, but it's the same person, or at least her twin. Cool, right? So that is the power of using
the seed parameter. You can only get the seed number from images that you create. You can't get the seed number
from other people's images. So don't try to clone someone
else's girlfriend, okay? Clone, only your girlfriend. Are you having fun? Great. Creating AI images is
supposed to be fun. Here's another example
of using seed, except it's with
our green dragon that we created earlier. This is the original photo. If we send ourselves
an envelope, we'll get the C number
for this image. And if we wanted
to, we can change the prompt to red
dragon in the forest, using the S number of
this original photo. Want to see how it turns out. Here we go. A pretty close selection to the original
green Dragon, right? Using this seed parameter is pretty helpful when
you are trying to create more consistent images of the same character for a
children's book, for example. Remember, AI is
random, sometimes. So if the image doesn't come
out the way you wanted to, redo it several times, and it should be somewhat closer than if you didn't
use the seed value.
10. 09 Consistent images with Reference Photos: One of the most powerful
aspect of creating an image is not just a stand alone image,
but consistent images, creating images that are like other images
so that you can turn it into a book or tell a complete story
where the topic, the subject, the character, stays the same, and it's
consistent looking. The best way to create consistent images is to
use a reference photo. And it is really easy to do. You can first drag your
picture into mi journey. Second, left click on the image to highlight
it and enlarge it. Third, right click the
enlarged image and select a copy image
address and fourth. Pays the link
before your prompt, which will often be a
description of the picture, as well as other words that give additional details
such as styles, themes, perspectives,
and artists. So, let's do this
together, shall we? I have a photo of a Batman mask. I'm going to drag it
into mid Journey. Then I'm going to left
click the image and then right click and select
Copy Image address. Now I have the location
of the image copied. I'm going to go into
the prompts section, it slash Imagine and copy and paste that link
into the prompt field. And then I'm going to say
that this is going to be Batman mask in a Museum surrounded by people
admiring the mask. And we get this. Not exactly
like the mask that was used, but we are heading in
the right direction. Notice how it's a mask of
Batman, not Batman himself. So even the bottom right
photo looks more like a Mannequin than
the actual batman standing in the museum. Here's another way
you can prompt AI while using a
reference image. Use the parameter C REF, and then the link to
your reference image. CREF stands for
character reference. You can see in this example,
we have the prompt. And then at the
end, we have CREF, and the reference image. Notice how sharp and high resolution the
reference image is. For best results, use a very
sharp and clear picture. If you use poor quality photos, and it's a bit blurry or so, it will give you a
very bad output. Because remember, high
quality in, high quality out. Bad quality in, bad
crap will come out. In this example, we
also used CW space 20. CW stands for the character
weight parameter. The value ranges 0-100, where CW zero focuses on
the character's face only, and higher values uses the character's face,
hair, and clothing. The prompt is a teacher
hugging toddlers in the classroom in a water
colored children's book style, delicate pastel colors. CRF, and the link to
the reference photo. We get this, which is very much consistent with the original
reference photo, right? And when you upscale one of them and just compare the two, looks like it was drawn
by the same person with the same material in the same style and
the same character. Cool, right? Okay, let's do a couple more
examples so you can see what's possible when using reference images to
create AI images from. Let's use the famous
fol Tower from Friends. A very simple and clear
picture is always good. And the other is Brad Pitt and Angelina Jolin when they
were still together. Who is sad about their breakup? What? Just me, really? Just me. Anyways, that's our reference images for today, whether you
like it or not. Let's use a certain
style for both. The Russian painter,
Alexi Von Jolinsk, who is known for
his bright colors and exaggerated bold lines. For the Iffel Tower, our prompt
is just simply linked to the reference image painted
by Alexi Von Jolinki. This is what we get
with faces randomly inserted in because
Alexi mainly do faces. And here is the same style
with Brad and Angelina. Up next is a vintage style. The prompt is linked to
the reference image. The Iffol Tower, vintage, detailed, intricate.
So what do you think? It is in a vantage style. AI created a border
for two of them, and a little castle underneath. I guess Napoleon lives there. Here's the same style with the movie couple, vintage,
detailed, intricate. Once again, don't freak out when AI gives you something
super weird, like turning Angelina
into a bird. I don't know what AI is
trying to tell us there, but the third one looks
pretty good, right? Kind of reminds me of Leonardo Dicaprio and Kate
Winslet in the movie titanic. Very classy and
glamorous looking. New prompt is linked to
the reference image, the FL Tower vintage
abstract expressionism. S 750 for a very high amount
of stylized creativity. And C 75 for a large amount of chaos freedom for AI to change the image. If you forgot these
parameter adjustments, don't forget to
watch the previous episode where we covered that. So, what do you think?
Creative, right? I particularly think the
floating tower look is very artistic and symboltic,
almost like a logo. Then we have Brad and Angelina. Yes, it gets really
crazy because we give AI a lot of
freedom to experiment. This is when the parameters that we set is way too powerful, and we would tone
it down a bit and continue experimenting with
the style, if you like it. Next prompt. Link
to reference Image, the FOL Tower
Futurism, Future Punk. S 250, toning the style lise
the creativity down to 250. Now, it looks like someone rebuilt a tower to make it look more space age or added
some modifications to it. One of them have a
futuristic fare going on. And the same happened
to Brad and Angelina. Looks like they are
in a space opera. If you like an image, even
if it's not quite there yet, don't forget to upscale them, because just by doing that, you will get more
resolution and details, and you may like the result. You can see the painting strokes from Alexi really comes out. Same with the movie couple. You can see the painting
strokes came out really well on Brad's cheeks and
Angelina's neck. Up skiing will let you see the details that you normally will miss when they are small. Remember I told you, I really like the floating Eiffel Tower. When you blow it up, it
looks great, doesn't it? Abstract with a smooth
gradient at the top, and a messy dripping
paint job at the bottom, as if the art was just created. Would you like to have
this in your living room? Well, with A ya,
you totally can. And who could forget the Titanic movie if Brad
and Angelina was in it. Really detailed
and three D like. Here is our futuristic
eiffel tower upscaled, looking more like an
electricity filled Tesla coil that can shoot electricity
than a harmless tower. Remember, you don't
have to go heavy with styles to make the
reference image look good. You can use very simple
prompts like this. Linked to the reference
image, a couple in love. S 500, C zero. And you'll get this, a lot
of stylization and no chaos. One of them looks like Brad Pitt turned into Robin Williams. But other than that, because our parameter has a
lot of style value, we get some pretty
good artistic choices. Look at the shadows.
Amazing work. When you turn the style
down and the chaos up, we get dis prompt, linked to the reference image,
a couple in love. S0c 75. You'll get this. So you might be wondering, why would I turn the
stylized parameter up and down or the chaos
parameter up and down. So you'll get images
that are more creative. You may start off your project, wanting to duplicate
Brad and Angelina. But then in the middle, you may realize that, Hey, I think Brad looks
great in a heavy beard, or Angelina looks great with
curly hair. Like this photo. Doesn't it look like it
belongs on a romance novel? That is where you get good
inspirations from AI. So over time, you
will get used to and more comfortable using
more different styles, keywords, artists names,
photographer names, aspect ratios, chaos, and more. And then you will really
be an amazing AI artist.
11. 10 Consistent images with Guided Images: So far, we have learned
how to prompt AI by using a specific
prompting structure. We talked about using different styles,
different variables, and parameters to fine
tune the picture, and we also covered
how to generate specific images from
reference images. Today, let's talk about
framing the picture, which is a technique in
me journey where we get a picture by giving examples
of images that we want. It's also using
reference images, but with an intentional angle, size, and even text in
the original image. Now, if you have used mid
Journey to create text, you'll know, it
is not that good. It is actually really crappy when it comes
to creating text. For example, here is the prompt, a photo of the Las Vegas sign during a snowstorm at Sunrise. And we get this.
But notice that, although the sign
looks pretty good, the lighting the
angle, it's all good, but the text is completely
wrong. It's gibberish. L as Vad favsa L a vaga Laia, which is the name
of Las Vegas in maybe another dimension
or another universe. In our universe, if you want a sign that looks closer
to the actual Vega sign, we can get that by framing the picture or by
using guide images. Just like using this
reference image, upload it into mid journey, and then copy and paste
the image location. And we can use this
as a guide image, as well as turn down
stylization and turn down chaos to guide AI to creating
an image like this one. C haos, remember, is how much freedom you are giving AI to stick
to your prompt, and stylization is how creative or surreal
you want the image. If we truly want to
guide our image, we will turn those
two parameters off. So our prompt will
be reference image, a sign for Las Vegas, C zero, S zero. With that guide, now AI is able to come closer
to the actual sign. Notice how it is still
not quite perfect. There are some typos, but these can easily
be fixed in photoshop. And this was done with chaos
and stylization set to zero. Since the original
reference image, the guide image is y screen. We can change the aspect ratio to push AI to create
an image more like it. This time, the prompt
will be reference image, assigned for Las Vegas during
a snowstorm at Sunrise, c0s0, AR, which is the
aspect ratio, 16 by nine. And we got this. Now when
we scale one of the images, you can see it looks very
similar to the guide image. With photoshop, you can
definitely do a little touch up to get to the final
result we were looking for. Here's another example of
using a guided image to help guide the AI into the
direction you want to go. One of my clients took a picture of her
daughter like this. She is bending over, holding
onto a yellow plastic stick, pretending it is a flower. Using this as the guide image, we can use the prompt. Reference image,
African American girl in a yellow dress smelling
a beautiful sunflower, comic art, Pixar cartoon. And you can see, AI changed the yellow plastic stick
into a actual yellow flower. Also, in one of the pictures, added a row of yellow
flowers in the background to make it as if that's where the girl got
the flower from. AI is also very good at
blending the colors together, like the flower,
like the background, like her dress, to make
it pleasant to look at. Once you find a style you like, make sure to use
the same variables at the end of each prompt, like comic art,
Pixar, and cartoon. You will use that over and
over again to continue to get consistent looking
characters for your project. This is a reminder that
you don't have to have an exact picture of what
you want to create, but simply using it as a guide, which is framing the picture. Here's another
example. A husband wanted to turn his
wife into a warrior. So we took her photo
and use it as a guide. The prompt is reference
image, fierce female ranger, strikingly detailed
armor and shoulder pads, stunningly beautiful, eight K, intricate, detailed,
ultra clear photo. And AI was able to create
these images in mere minutes. So what do you think?
Pretty cool, right? Now this person's
wife is a warrior. Look at the lighting,
the shading, the details, and the textures. Imagine creating a calendar
for this person, a book, or some social media
promotional images for this person's business
or influencer promotions. How much would this
person pay you for all of that work
done in minutes? That is how you can start
thinking of using AI, not just for yourself, but to create multiple
income streams.
12. 11 More examples of Guide Images to make consistent images: The two most popular types of consistent characters
you will be creating for your projects are usually comic styles or
children's book styles. Now, comic styles
are harder to create consistent characters
for because there are much more
realism involved. As you add more
and more details, it makes sense that
it becomes harder to make each character
consistent, right? The children's book
style is a lot easier to create because the images are
just quite simply simpler. And sometimes, when you
add a specific style like watercolor or oil pastel
or surreal style, it actually
simplifies the image, and consistency becomes easier. Here's an example
of a comic style. One of my clients is
creating a comic book, and it is about
the adventures of a Green elf girl in
a fantasy world. Now, taking that
as a guide image, and creating multiple versions
of this, we get these. Now, my client, just
for a disclaimer, created at least 150 images from the guide image.
Let me repeat that. My client, based on
the guide image, created over 150 extra images, and these are the ones that
came out looking the best. So as a sino, just keep in mind, you do have to create a lot of these images to
be able to have a handful of them that you will be satisfied to
use in your project. Even with that, you can see not all images look
exactly the same. The character does
have minor variations. Sometimes her hair
is worn differently. Sometimes the mark on her
forehead is different. And sometimes she is
wearing different wardrobe. And we, as storytellers, can get away with that. Why? Because a character doesn't always wear the same
wardrobe in every photo, or look exactly the same, right? It would be boring if
that was the case. So it is okay if the photos
are not exactly the same. And these were all created from one guide image, the
reference image. Now, let's talk about
creating letters and logos in my journey
with a guide image. Remember how AI is not good at all at creating
text in the image. Here's a good example of that. Let's say we put in the prompt. An intricate logo with
the letter H, eight K, intricate, detailed,
ultra clear photo, hyper realistic, complex
photo realistic. And we get this image. It's cool and pretty, no doubt. But all of it is more of the letter B than the
letter H. Isn't that weird? AI is really smart, but sometimes it just
doesn't know its alphabets. Now, what will happen if
we help guide the AI with a guide image of N H.
Here's our example. We wrote the letter H with the impact font because it
looks really bold and clear. And we put this into
mid Journey and asked it to create that as
the reference image. Using the same prompt as before. This time, with a
reference photo first, we will get reference
image link, an intricate logo with the
letter H, eight K, intricate, detailed ultra clear
photo, hyper realistic, complex, photo realistic.
What do you think? You know how there's a
lot of people online complaining that they can't
get letters in their images? Well, you can, as long as you keep trying
and guide the AI. Then, just for fun, I did an
experiment to see if we can get away with using
three letters without using a guide image. This time, we will use
a very long prompt. Asking AI to create a really
fancy image with texture, lighting, style, and et
cetera. Here's the prompt. Don't freak out, okay? It is simply a large
sign that says LOL, and then AK,
intricate, detailed, ultra clear, photo, hyper
realistic, complex, photo realistic, multi universe scattering glow
shadows, and et cetera. Guess what we got. The
letters are indeed fancy, colorful, detailed, stylish,
stunning to look at even. But what the hell
happened to LOL? AI decided it's
gonna make EOI 80, another 80 and a
AO 19 super weird. But what if we added a
reference image like this, with the text
clearly spelled out, L O L. And with the large
prompt that we had before, we will get this. And if we up scaled one of them, it is exactly what
we were looking for. And then we can try to
fine tune it and add colors and add
details and texture. But that would be how to
create text using M journey. Are some important tips when creating consistent characters. One, always decide
your style before going for consistent characters
or scenes in your book. Two, generate many more
images than you need. Many can be cropped or lightly edited to
make them perfect. Three, always upscale
all four images. Many times the upscale will fix small errors in
the first draft. Four, Always do variations
on images you like. And sometimes a
lucky one will come through so you can
upscale. And five. Consider using a lesser known
movie star or celebrity as inspiration for a character or even a mix of two
or more people? For example, if you are creating a children's book with
a child character, if you use McCully Cogen, probably everyone is gonna
recognize him because, hey, he is the Home alone boy. Who can forget him. But do you know the character
Punki Brewster? Probably not, right?
Because she was a character in a TV
show back in the 80s. Most people will
not remember her. So if you create
a children's book using her in the prompt, like a young redheaded
girl on a swing set, punky Brewster, watercolor
style of Beatrix Potter, you get this character, which is a great start. And here is her crying. And here is her
eating ice cream. Notice how the character and its style are pretty consistent. This is why it's important
to always decide your style before going after
consistent characters or scenes in your book.
13. 12 Advanced Character Creation Part1: Hey, my friend, you made it. Welcome to the Advanced
character creation, where in this episode, we will cover
advanced ways to get consistent images for a
character that you are creating? Roll up your sleeves because we're gonna have
tons of fun today. We talked about using guided
image or reference images to help guide the AI in the direction you want
your images to turn out. And I bet some of you
may be wondering. But, June, where do we
get the guided image? Or what's the best way
to create a guide image? When you want to get a lot of guide images of the
same character done. There are numbers of prompts you can add into
your description. These will really
help you create a character and keep that
character consistent. These prompts
include, but are not limited to expression sheet, list of expressions,
character design sheet, full body reference sheet, character reference sheet
with several angles, Isolated design asset,
full body view, side view, rear view, T pose, front and back view, two character sheets of X, front view, back view,
and character model. For our example today, we will talk about a
character that one of my clients created,
which is a Blue J. Here's an important
thing to note. If you're starting from
zero and a complete beginner trying to create
images for a children's book, for example, it is
highly recommended that you use animals
as the main character. Why? Because animals
are a lot easier to create consistent
images for than humans. So let's start with the first
example prompt, shall we? L large blue J, multiple
poses and expressions. Children's book illustration
style, simple, cute, blue, yellow feet, white crest, perfect beak, flat
color, and no outline. Notice how the beginning
of the prompt, the texting blue will change
as we create new images. But the last part of the prompt, the one in black text will stay the same every time
we create an image. Why? Because that is a style
that we have decided upon, that should not change
from image to image. And here's the mock up for
this blue J character. Notice how in each photo, you can see the feet. That's because it's
part of our prompt. We told AI it has yellow feet, and that's why it included it. Once we upscale the image, it can really help you decide what direction
you want to go in. Like this one, if you bought this book for your child,
how would you feel? You'll probably wonder,
why is this bird so angry? Looking, right? It's like
this bird is in a bad mood, which is probably what you
don't want in a Kids book. And you can see another
upscale mockup. This came out pretty good. But depending on
your preference, you might feel there's too
much yellow on this bluebird. And because the title of
the book is Blue Bird, not blue and yellow
bird, at the end, my client went with this one as the default character design. The default guide image
that will help push AI in the right direction
every time you work on this character. Here's a tip. To use this mockup effectively, it's a good idea to crop and create a separate file
for each of the bird. So that you have one image
looking slightly left, one looking left at
a 45 degree angle, and another looking
completely left. And you can crop them using any photo editing app like
even Microsoft Paint. When you do this, you now
have many references guide image for this character with multiple poses
and expressions, depending on the angle, the pose, and the expression, you can pick the guide
image that you want and create images based
on that reference photo. Like, let's say,
the first chapter of the children's book is the Bird Happy on a tree
branch. And here's the prompt. Reference Image link,
large Blue Jay, happy on a tree, branch, joy. And then the same
list of styles and variables and parameters
we covered earlier. So with this as our guide image, we are able to create
these images from that. What do you think? If you are just starting from scratch
with no experience, this is a great start, isn't it? Well, we can continue. Using this as the guide image. We can write another prompt. Reference image
link, large Blue J, angry, side on side view,
chirping at a child. And we get this. It
got the first part, right, but it's not
chirping at a child. And if we really wanted that, we can add weights
to the prompt. Like reference image
link, large Blue J, angry, column column one, space, side on, side view, column column two, space,
chirping at a child, which will make the side
view and chirping at the child more important
than the angry Blue J. It is also important to note that if you hit the redo button, it will not include
the seed image. To use the seed image again, you need to copy and paste
it into the prompt again each time you want to use
it. Here's another prompt. Reference image link, large
blue J sitting on a branch, left view, left looking. We get this with at least two of them
sitting on the branch, looking towards the left. And when we upscale them, the image will look
even better, right. Now, the eyes and the
detail are more prominent. Let's actually do
awaiting experiment. Using this guide image, looking straight at
us and this prompt. Reference image link, large Blue J standing on
a picnic table in park, Col Column one, a
large park background, Col Column two, meaning the large park background is more important than the bird. And then we get this.
What do you think? It definitely captures
a child's imagination because the background
is very prominent, and with the bird front
and center facing us, it looks like a gigantic bird
and the hero of our story.
14. 13 Advanced Character Creation Part2: Now, let's create
another character so that I can show you more advanced examples of using a model sheet as
your guide image. Remember, these are
the model sheets that I recommend you
experiment with. This time, let's create
a child character. Here is the prompt. A
young red haired child, multiple poses and
expressions model sheet? Children's book illustration,
style, freckles, red hair, white shoes, blue shorts, flat
color, no outline. And AI gives you this. Because you didn't specify
the gender of the child, some of them are boys, and some of them are girls. Then you can upscale
the boy model sheet and the girl model sheet
to decide whether you want the character
to be a boy or girl. Isn't this creative and fine? Before AI, you would
have to hire an artist, and they would have taken hours and then
gotten back to you, and then you will give
them some feedback, and then they'll make
the adjustments, which will take
hours, if not days. And after that, get back to you. And back and forth
back and forth, w is in seconds with just
you and your friend, AI. Let's say we like our character
to be the little boy. Then we'll do the same as
we did with the Blue J. Crop each of these characters
into separate files so we can train AI on what
direction we are going for. Like, we can use this as the
guide image and the prompt. Reference image link, a young red haired child
waiting for a bus. And we get this, would you
agree? Is pretty close. Now, using this as
the guide image where the child is
sitting and the prompt. Reference image link, a
young red haired child sitting in a classroom with
other children and a teacher. What do you think? Yes, I know. The one pick in which
she is sitting on air, because the chair is
missing is pretty funny. But you get the idea, right? Two of them still
look like a girl, because we have yet to
specify the gender. So in the future,
our prompt can be a young red haired boy
sitting in a classroom. Would you agree that
each of these images help you tell the story
of your character, and the audience would feel
it is the same character. Want to discover another hack in creating consistent images. You can create a model sheet based on your reference image. Yeah, clever, right? Let's say you like
this model sheet. Well, you can create a model sheet using
this model sheet. I'm not kidding you.
All you have to do is take this image and drag
it into mid journey. Get the reference
image location link, and use this prompt. Reference image link, a
young red haired child, side view, rear view, multiple poses, model sheet. And you can create
multiple model sheets from that one model sheet. There are many different
kinds of sheets. Here, we have an expression
sheet where you can get an idea of the different
expressions of your character. And the prompt to
get this is simple. It's just reference
image link, a young, red haired child,
expression sheet, angry, happy, sad, scared. Here is a full model reference
sheet that will show your character's
body proportion in different angles. Don't forget. Here is a list of different sheets you can experiment with. So as you create more
and more sheets, you can simply crop the
ones you really like and make more images or more
model sheets from them. Cool, right? Now,
let's talk about using two reference images
to help guide AI into creating
an image for you. Here, you can see our
reference image one. And here is reference image two. This is where we
are trying to add both the child and the
eiffel tower into one image. So our prompt will be
reference image L link one, reference image link two, a young red haired child. And you can see, we
get one image where the eiffel tower is kind of
visible in the background. If we now prompt the AI to use both reference image links and a young red haired child standing in front of
the Eiffel Tower, we get a better image. When I was doing a bunch of
experiments, for one prompt, I forgot to add that it's a
children's book illustration. And I got this, which made
it look more realistic. So it is important in your prompt to always add
the style, variables, and parameters at the
end because those don't change if you want
consistent looking characters. Remember, to also experiment with both the reference image and the seed to get as
consistent as possible. Like, to create
this model sheet, I use both the
reference image of the model sheet plus the
seed at the end with the d, d, C number of that model sheet, to really do everything
I can to force the AI to come up with
consistent images. So when you look at
this entire prompt, you have the
reference image link, the specific instructions for that prompt, the same style, variables, and parameters
that don't change, and the seed number at the end, all working together to guide the AI to create the image
in the direction you want.
Jun Wu, Social Media Influencer
