Workflows

Character series

Train once, reuse across renders.

Aura ID lets you train a tiny LoRA on a single subject and then @character-name that subject into any generation prompt. Ships with M6 β€” today the training surface is stubbed at /create/character, but the concept:

  1. Upload 20-40 reference photos (multiple angles, consistent lighting).
  2. Name the character + attach portrait-rights consent (PDF).
  3. Submit β€” H100 training runs 3-5 minutes.
  4. Use @name in any /create prompt β€” same face, different scene.

Your LoRA stays private to your workspace. You can share it with specific members once workspace teams land in M8.

Last updated 2026-04-17Needs more detail? β†’