Beta
Docs
Modes

Avatar Generation

Create character portraits, full avatars, and flat-lay outfits with identity and reference controls.

Avatar Generation builds character images with structured inputs: identity, outfit, environment, and an optional prompt builder. It is not limited to a single face-swap style — three input modes change which slots and models you use.

For pricing, see the Pricing page. Models: Image Providers.

Input modes

Avatar ComposerFull avatar with identity, outfit, environment, and optional contact-sheet grids.
PortraitClean white-background portrait driven entirely by the avatar builder — no media slots.
OutfitTop-down flat-lay outfit generation with all available outfit reference slots for the model.

Composer extras

  • Production sheet — optional multi-view character sheet when enabled
  • Grid layout — contact-sheet style multi-tile outputs (disabled when production sheet is on)
  • Style reference / preset — optional style lock when custom style is enabled

How it works

  1. Choose an input mode (Composer, Portrait, or Outfit)
  2. Fill the slots that mode expects (or rely on the avatar prompt builder)
  3. Pick a provider and model (the list filters by mode)
  4. Set aspect ratio and quality as supported
  5. Generate

Providers (by mode)

Composer and Outfit need models that can take references; Portrait can use broader image-gen models.

ModeTypical providers
ComposerOpenAI, Gemini, Seedream (reference-capable avatars)
PortraitBroader image-gen set (Gemini, Flux, Qwen, Seedream, OpenAI, Recraft, Ideogram, Krea, …)
OutfitGemini, Kling, Wan, Grok, OpenAI (multi-ref outfit slots)

Exact model lists match the Create picker for Avatar — see Image Providers.

Tips

  • Composer works best with a clear identity photo plus optional outfit / environment
  • Portrait is ideal for consistent headshots from builder attributes alone
  • Outfit mode: fill as many flat-lay references as the model allows
  • Use the Node Editor when you need batch or pipeline avatar work

On this page