What are Styles?
Layer comes with multiple pre-trained styles that you can use to create art assets. These styles were created to show the possibilities with Layer using our own assets used for training purposes.
Think of styles as aesthetic art styles, such as pixel art, anime, digital painting, cartoon, etc. Layer is able to support any art style as long as it is provided with sufficient training data. You can train your own if you are on Pro and above plans.
Does Layer only support cartoonish style?
No. Your input defines your output. Therefore, if you like to create assets that are more realistic, you will need to train Layer with such assets to create your ‘realistic’ art style. The same rule applies to any style you would like to create on Layer.
How many styles can I create?
All of our paid plans except Free allows you to create custom styles. The limits on number of styles you can generate on Layer varies between plans.
Which training resolutions are supported?
512x512: Styles trained with this resolution will be faster to forge yet lower in visual quality and text comprehension compared to 1024x1024.
1024x1024: Great for use cases like characters and environment, background assets and is generally a superior option when training styles. One drawback is it's slower to forge with.
Creating a Custom Style
Since your project has a distinct art style, the default pre-trained styles might not match the aesthetic your assets need. This is when you’ll need to create a custom style (also known as training a style).
It’s encouraged that you will likely have multiple custom styles per game project. To get the most accurate generations, you should train a separate style for each part of your game, as well as a general style that you can utilize to experiment with new concepts.
So, for example, let’s say you have an RPG game that looks like this:
You might have styles for the following:
This style would be used to generate new concepts or explore ideas for a game while still adhering to the overall style.
Background / environment style
Inventory item style
You can see from here that more general, broader styles are better for concepting and exploration, while more specific custom styles are better for production assets that have to be more specific.
In addition to the above-mentioned use cases, it's important to emphasize, you can train separate styles for various other use cases such as terrains, textures, maps, buildings, props, animals, icons, badges, plants, weapons, clothing, drawings, decorative items, illustrations, paintings, even photography!
Creating a style, step-by-step...
1. Select Your Training Resolution
Layer currently supports two resolution options. 512x512 and 1024x1024.
2. Select your use case
Single Character: For creating poses, angles, and varying shots of the same character.
In-Game Items: For creating items and equipment like trophies, gems, weapons, etc.
Backgrounds: For creating game backgrounds, be it matte paintings, isometrics, etc.
Multiple Characters: For creating variations of a character type. E.g. avatars in the same style.
Vehicles: For creating multiple devices and vehicles. E.g. variations of cars.
Environmental Objects: For creating variations of environmental objects like houses and trees.
UI: For creating frames, buttons, and more in your game's UI style.
Icons and Symbols: For creating icons and symbols in your consistent art style.
Other: None of the above.
3. Upload your images
The first step in creating a custom style is to gather 6-25 source images.
These source images will be the basis of how Layer creates a custom model to fit your style, so it’s important to select the right ones for best results.
4. Caption your images
Writing captions are found to be non-trivial by most of our users that's why Layer comes in handy with auto-captioning feature which auto-populates captions for the images you uploaded and make it easy for you to review and make changes were relevant.
Captioning your source images is just as important, if not more important than the images themselves.
When captioning an image, you want to be as descriptive as possible, as if you are describing the image to a blind person.
Use commas to separate descriptions.
If some of the characters or objects have proper names, you can also include them so that later during generation you can just refer to it by name and the model will know what you are looking to create.
5. Add evaluation prompts
You’ll also need to pick some test prompts. These text prompts will be used during training to generate sample images so that you can evaluate how the custom style is doing.
We’d recommend at least 3 sample prompts that cover a range of intended use cases. For example, if you have a custom style that is meant for inventory objects, you could the following example prompts.
A silver sword with a golden hilt and a red jewel on the tip of the handle.
A red healing potion with a cork and bubbles inside it
An ancient scroll with runes written on it tied with a frayed blue rope.
6. Wait until your style is ready!
After a few moments, your style will complete. You will receive an email when your style is ready. At this stage, you can start forging with it!
7. Select the best style variation
Once Layer completes training a model for your custom style, it will generate several variations. When Layer trains a model, it takes several guesses at what might have the best results and picks one for you, but that means you can also manually go in and inspect each variation and see if there is one that is better.
Click on “Style Variations” to see your custom style’s variations. For each of the sample prompts, a sample image will be generated in the custom style. Pick the one that you think works the best across your prompts.
8. Describe your style
Write a quick description to make it easy to for yourself and other team members you are sharing your styles with. You can also add tags which will help you search and find your styles on Layer.
9. Add negative prompts
You can add default negative prompts for Layer to respect every time you forge with the style. Note that you do not need to re-generate the style after adding these for these to be incorporated.
EXAMPLES OF COMMONLY USED NEGATIVE PROMPTS:
ugly, 2 heads, duplicate, blurry, abstract, disfigured, deformed, cartoon, 3d, disfigured, bad art, poorly drawn, extra limbs, close up, b&w, weird colors, blurry, watermark, blur haze, watermark, elongated body, cropped image, out of frame, draft, deformed hands, twisted fingers, double image, malformed hands, multiple heads, extra limb, ugly, poorly drawn hands, missing limb, cut-off, over saturated, grain, low res, bad anatomy, poorly drawn face, mutation, mutated, floating limbs, disconnected limbs, out of focus, long body, disgusting, extra fingers, gross proportions, missing arms, mutated hands, cloned face, missing legS
Forging using your style.
After completing your style, you can use it to create assets using prompts.
Prompts are written instructions to help guide an AI model on what asset to generate. However, just like learning a new language, there’s a specific way to talk to the AI to get the results you’re looking for.