Any hints on how to not always generate
the model type? I want a normal person. Not a supermodel, not beautiful, not ugly.
Just an average face. Well izzmot you're right that Midjourney likes to generate
images that look like Photoshopped supermodels. To generate photos that look more
normal we can use negative prompting. First, let's start with a prompt like
"portrait photo of an average American young woman, everyday life". A place where
people might post day-to-day photos would be Facebook, so let's also add that
in there. Most importantly, we need the Negative Prompt accessed by using the --no parameter. Here we'll write down all the stuff that we don't want to see in
the generated image. Words like model, makeup, photoshop, and perfect. I also use
Raw Mode to minimize Midjourney's creativity. And now we can easily
generate normal looking people that don't have the super photoshopped model
look. I get asked a lot of questions like that
in my comments section. I think a lot of them are common practical problems that
you'll encounter when starting out with Midjourney. So let's walk through some
of the solutions to the more common problems today. Hi, why does Midjourney
produce distorted faces, nose, or hands while generating images with multiple
characters. I've tried a lot but haven't been able to produce decent images with
multiple character characters please guide. All right so the smaller the size
of the face relative to the rest of the image, the more distorted it'll be when
Midjourney generates it. This is especially true if you're doing images with
multiple faces. Take a close look at these hikers in the mountains. Their
faces are totally warped and unrecognizable. One tip to try is using
the Creative Upscale feature which upscales your image, but also makes some
subtle changes along the way to improve image quality.
I find that most of the time Creative Upscale fixes the distorted faces when
you have multiple people in a photo. Take a look at the before and after and it's
night and day. The faces look way more human after upscaling, although that hand
is still kind of weirdly placed. Here's another example, look at
the girl on the right side of this image. Her face is really blurry and doesn't
have the silhouette of a person. After using the Creative Upscale
though, we can clearly recognize that it's a woman with glasses
on. Hi thanks, could you please provide a prompt best suited for t-shirt mockup
photoshoot. Alright so check this out. The explore homepage on the Midjourney's website features images that millions of other Midjourney users have generated.
This is the best place to look for inspiration if you need specific prompts
like mockups. In the top right there's a prompt search bar. Let's search for "shirt
mockup". It looks like there's a ton of other people also generating t-shirt
mockups. I really like this one that has a forest background in it. It looks very
outdoorsy. When we hover over the image, there's this icon that shows a capital T
inside a square which is the edit prompt feature. Click on this to copy and paste
a prompt into the prompt message bar at the top. I'll press enter to run the
prompt and navigate to the create tab on the left, where we can see all of the
T-shirt mockups being generated for us. I think any kind of outdoors, hiking type
of branding would look really great on these. Back to the original image, you can
also use this search image feature that looks like a microscope and Midjourney
will find other user generated images similar to it. Almost all of these have
black t-shirts with some nature forest theme in the background. If you don't
have access to all of the features of the new Midjourney website yet, you can
always just copy and paste the prompt from the image and then use it in
Discord. I think you should make a video on the insane new style reference and
how complete Brands can be created just from concepts now with style fluidity.
I'm working on a portfolio showcasing how insane the potential is to show
clients who could use full product lines created all in Midjourney. That's a
great video idea, let me test it out. How we can use a style consistency feature
to create a complete product line. First, I need reference images. I'm going to
head to pexels.com, but you can get these images from anywhere. Now I want a rustic
old school kind of vibe, so I'll search for a rustic vibe in the search
bar. Something like this photo of a countryside with a wooden cabin looks
perfect. Let's download it. I finished collecting five images that I want my
product line to be inspired by. They all have this countryside farm, rustic kind
of look. So let's head over to Midjourney and start generating our
products. I'm in the create tab here. I'm going to use the upload button that's
the plus icon to upload reference photos. Let's just drag and drop our rustic vibe
reference photos in here and then we'll attach them into the prompt box. When we
hover over the image we see three different icons. The person icon is for
character consistency, the paperclip is for style reference, and the photo is to
use it as an image prompt. We'll be using the paperclip icon for style reference,
and let's go ahead and do that for all five of the images. Now inside the prompt
bar I'll ask for something like "product photography, a bar of white soap, there's
a minimalistic rustic background." This should give us a product photo of a soap
but inspired by our five reference images. And the result definitely has
that handcrafted outdoors rustic mood going on. I tried out some other products
like a bottle of perfume on a rock, this candle on a log, some winter boots. I like
the pine cone decorations here. Notice how they all keep the same style and
look like they come from the exact same brand. Hey bro, hope you reply. My question
is that all the things you're trying are basically normal and somewhere we've
seen them. That's how it feels for every image. Here what I want to know is that
there's different groups of tribes, what I'll get on Midjourney is some African
and some other, but I want totally different culture, different tribe looks.
Their looks should feel like the world hasn't seen these type of tribes. Is Midjourney capable of generating these kinds of images. So you want Midjourney
to generate something unique that we haven't seen before. If we just ask for a
simple prompt like "photo of a ancient tribe of people". Midourney doesn't have
enough information to create anything unique, and will just give you a generic
looking photo like this. We'' have to use our creativity and give mid Journey
some more clues about what we want. Let's start with "full body shot of a tribe of
European people with blonde hair and light blue stripes of paint on their
faces". Using our imagination we can ask them to have yellow pupil in their eyes.
Now it's already something unique, but I'll go even further and change up the
clothing they're wearing to give them giant blue and yellow
feathers. And finally they should also have weapons made out of bones. I've
never seen images of a tribe like this. It's definitely something that hasn't
existed before. Bro are you real or AI? Yes....... I am Please can you create a tutorial on how to create images like the AI Historian
or Stellar Sagas? I think you're talking about these AI history videos and shorts
that have been going viral everywhere. If you want to figure out how to prompt for
specific images you like, try using the /describe command in Midjourney. Type /describe into the direct message and then you can upload a reference image.
I've taken some screenshots from Stellar Saga's AI history videos which I'll
upload here. I still have no idea how to use the /describe command with the Midjourney website so let me know if you figured that out. When you use this
command Midjourney generates four prompts that roughly describe the
reference image you uploaded. These aren't supposed to generate that image
exactly, but it should give you some hints. I'm going to hit imagine all to
run all the prompts and some of them look really similar to the original reference. We can go a step further and
combine the /describe feature with style reference. Here I've described an image
of Mongolia soldiers on horseback. I'm going to copy one of these prompts and
paste it into the prompt box. After that I'll use the original image as a style
reference. So right click and copy the image link, and then inside the prompt
type --sref to use style reference and paste in that image link. And I'd say we
ended up with a picture pretty close to the original style. Quick question, for
general PowerPoint background images generated by Midjourney, which general
camera type, film stock, and camera lens do you recommend? That depends on what
kind of PowerPoint you're trying to make. Adding the keyword stock photo at the
start of your prompts will generate generic stock images that look like
these. Nothing too fancy but it'll get the job done just
fine. Clipart gets you fun cartoon illustrations. These would fit in pretty
well on some PowerPoint slides that's more light-hearted and
relaxed. You can try Vector illustration and use the keyword "minimalistic" inside
the prompt for these abstract but clean and simple designs. May I ask how to arrange
multiple subjects in a scene? Yes, you may. For example, a girl is talking to a boy
downstairs from a second floor balcony. With a girl positioned in the top right
corner of the frame and the boy in the bottom left corner. These subjects are
not aligned on the same horizontal or vertical line. It's super hard to get
that kind of fine grain control in Midjourney using just the prompt box, so
we'll have to try some other methods. Starting with this image of a building
with a balcony I generated, I'm going to use the Vary Region tool in the bottom
right. I'll select the part of the balcony where I want the girl to be. That....
looks about right. Then inside the prompt add in "a young woman is looking over the
balcony". With that girl generation done, I'm going to use a Vary Region tool again,
this time selecting the part of the street where I want the man to be
generated. I'm actually going to delete the part of the prompt that talks about
the young woman because I don't want the man and woman to get mixed up by Midjourney. Next I'll add "a young man is standing in the street holding roses, he
is next to the balcony". So I tried this a couple of
different times, it's hard to get it to look very realistic. The faces are kind
of distorted and warped. Remember earlier I mentioned that the smaller the faces
are in the size of the entire image, the more distorted and warped they're going
to be. Even the Creative Upscale can't fix these since the face is so small.
Unfortunately generating realistic small faces is still a huge issue in Midjourney. Please, make one video on Niji version. Don't worry, I have some anime
prompt videos planned ahead. Here's a tip for now. When prompting for anime style,
use emotions and expressions to upgrade your pictures. Images with
powerful emotions capture the viewers attention in the way that images without
them just don't. Which one of these are you more likely to stop and look at? The
one on the left with a blank expression, or the one on the right of the girl
crying tears of joy. Take advantage of the expressiveness of anime whether
that's anger, or complete fear. Add some storylines to your photos for the
viewer to think about. This girl is terrified and running for her life from a
demon wolf. Thanks a lot for your content.
You're super welcome! Can you help me with the realistic texture of drenched
clothes? I created some rainy scenes and my subjects looks photorealistic but hair
and clothes are dry. Biggest challenge is also if you use different fabrics for a
rain scene or fight scene of warriors in a muddy landscape. Let's try to be extra
specific in the prompts about the texture the clothing should have. I'll
start with a photo of a Viking Warrior running on the battlefield. I guess cloth
would be a common material they used for clothing back then. Then I'll add that
it's raining and also say that his clothing is wet and covered with mud. His
hair is also wet with water and mud. At the end, I'll use Kodak portra
film. So I asked for rain but also specifically told Midjourney that the
clothing is wet and muddy, and his hair is also wet and muddy. I'd say that these
images look about right you can definitely see the mud and water on his
shirt. The hair also looks drenched with water. I also tried giving him chainmail
armor and the mud and water effects also look pretty good. It is true that Midjourney has trouble generating realistic water on hair and faces but if you
specifically prompt for it, the resulting photos will at least have the basic idea
down even if it doesn't look super realistic. Is it possible to have two or
more consistent characters in different scenes?
I've tried this using the /prefer_option_set for two characters and with two
character references at the end of the prompt and haven't had much luck. Well
Natalie, that's the most common question I've gotten. How to create multiple
consistent characters. To answer that, I've created a video tutorial over here.
So make sure to take a look if you're interested. Also make sure you subscribe
if you want to see more like this.
