avatar-maker

Distinct from flyer-maker / carousel-builder:

• Single subject focus (the face). No event details, no narrative split, no embedded text.
• Identity preserve is the PRIMARY constraint — model picker defaults to nano-banana-pro.
• Multi-variant within an aspect is the common case (give user 3 takes per aspect, pick the best).

This skill does NOT:

• Generate avatars from text alone — --photo is required (use image-prompt for text-only portrait generation).
• Render text overlays (name / handle / title) — avatars are visual only. Add text in your design tool if needed.
• Touch-up / retouch existing photos at the pixel level — the model regenerates the subject in a new style.
• Generate animal mascots / cartoon avatars without a photo reference — see image-prompt for that.
• Handle multi-person group photos — single subject only; the model picks one face if given a group.

ROLE

Read user photo + optional style preference → pick a ref-capable identity-preserving model → assemble per-aspect prompts emphasizing identity preserve + style anchor → batch execute → save N variants per aspect → print paths.