Stable Diffusion Prompt for Multiple Characters

Stable Diffusion is a powerful AI image generation tool that allows users to create high-quality images from text prompts. One of the most common use cases is generating images with multiple characters interacting. However, crafting effective prompts to control multiple characters can be challenging.

This article provides tips, examples, and best practices for composing Stable Diffusion prompts with multiple characters. Whether you want an action scene, romantic encounter, or just a group portrait, these prompts will help you achieve consistent, high-quality results.

Why Are Multiple Character Prompts Difficult?

When generating a single character, Stable Diffusion only needs to interpret one set of descriptive parameters from your prompt. This allows it to render features like facial structure, hair, clothing, etc. in a cohesive way.

However, when you add multiple characters, the number of descriptive variables increases exponentially. Without enough specificity, Stable Diffusion starts “mixing” character features. For example, outfit elements or hair might randomly swap between the characters.

The key is crafting prompts that provide unambiguous, discrete descriptions for each character. This article explains techniques to keep them visually consistent.

Describing Multiple Characters in Stable Diffusion Prompts

The first step is structuring your prompt to define each character individually. Here are some formatting tips:

Use Lists and Headings

Break characters into an enumerated or bulleted list. For example:

1. Lucy: 
   - Young female with red hair in braids
   - Wearing blue dress and holding a bouquet of flowers

2. Tom:   
   - Middle age male with short brown hair and beard
   - Wearing suspenders over a white collared shirt

Alternatively, put each character in a separate heading:

Lucy

Young female with red hair in braids
Wearing blue dress and holding a bouquet of flowers

Tom

Middle age male with short brown hair and beard
Wearing suspenders over a white collared shirt

Add Specific Physical Details

Include distinct descriptors for physical attributes like hair, skin tone, height, build, facial features, and clothing. For example:

Lucy: Petite female with fair skin, green eyes, and long blonde hair. Wearing a red sweater and jeans.

Tom: Tall muscular male with tan skin, brown eyes, short black hair and beard. Wearing a blue button-down shirt.

The more visual details you provide for each character, the less room there is for Stable Diffusion to “mix” features.

Setting the Scene

In addition to character descriptions, setting details provide critical context for Stable Diffusion to compose coherent images.

Establish Relationships

Explain how the characters relate to set up their positioning/body language:

A mother hugging her young daughter in a field of flowers. The mother has her arms wrapped around the daughter, who is smiling up at her.

Describe the Environment

Add information about the surrounding scene and props. For example:

Two wizards battling in a crumbling stone arena filled with fire and rubble. The first wizard is elderly with a long white beard, wearing purple robes. He is casting a lightning spell. The second wizard is younger with red robes, casting a fire spell.

With environmental context, Stable Diffusion can realistically integrate the characters and their actions.

Advanced Prompt Optimization

You can further improve multiple character prompts using these advanced techniques:

Create a Character Gallery

Curate a collection of reference images showing each character. Use img2img to fine-tune Stable Diffusion on rendering them accurately.

Generate Characters Separately

Compose individual prompts for each character, generate images, and composite them together in post-processing. This avoids “mixing” features.

Use Multiple Control Nets

Apply a unique ControlNet for each character to strongly guide Stable Diffusion during image generation.

Experiment with Sampling Settings

Tweak configuration like seed, sampling method, and steps to hone diffusion performance on complex multi-character prompts.

Example Prompts

Here are some full prompt examples for generating consistent images with multiple characters:

Fantasy Party Portrait

An intimate portrait of an adventuring party including:

1. Nora the human wizard: elderly female with gray hair in a bun, green robes, and wooden staff.  
2. Drek the orc barbarian: tall muscular male with green skin, long black hair, brown leather armor, and large battleaxe.
3. Pix the gnome rogue: short female with brown hair in pigtails, red bandana, brown tunic, and dagger. 

The party is celebrating their latest victory in a tavern, raising tankards of ale and laughing together. Detailed oil painting by Greg Rutkowski and Alphonse Mucha.

Superhero Battle Scene

Two superheroes battle a giant robot in a city street at night. Dynamic scene rendered in Marvel comics art style by Jim Lee and Jack Kirby.

1. Captain Light: athletic adult male with short blonde hair. Wearing a blue and yellow armored costume with white cape. Firing lightning bolts from his hands at the robot. 

2. Shadow Woman: curvy adult female with long black hair. Clad in a dark purple and black armored costume with face mask. Casting swirling shadow tendrils that grab the robot.  

The robot is 20 feet tall, humanoid, with glowing red eyes. It is knocking over a bus and shooting lasers from its arms at the heroes.

Conclusion

Crafting effective Stable Diffusion prompts with multiple characters requires clearly defining each one visually. Formatting tips like lists and environmental context guide the AI to generate coherent images. Dialing prompts helps overcome difficulties like mixed features.

With practice, you can create stunning, imaginative scenes with consistent multi-character results in Stable Diffusion. This opens up new creative possibilities for artwork, comics, game assets, and more.