generate-voice-portrait

name	generate-voice-portrait
description	Generate a 3-frame lip-sync portrait set for a speak voice and upload it to the daemon. Use when adding a new voice or refreshing an existing portrait.
allowed-tools	Skill, Bash, Read

Generate Voice Portrait

Creates the three portraits (mouth closed / slightly open / wide open) the dashboard uses to animate lip-sync, then uploads them to the speak daemon.

Paths below use {base} as shorthand for this skill's base directory, provided via the "Base directory for this skill" context injected when the skill loads.

Inputs

• name (required) — voice name, e.g. Josh. Used in the upload URL and to key portraits.
• style (required) — short descriptor of the character, e.g. "deep-voiced confident male, short dark hair".
• kind (optional, default default) — portrait flavor:
  • default → stylized portrait avatar, neutral lighting, shoulders-up
  • codex → robotic, terminal green glow, ASCII/monospace motifs
  • user → warm, human, friendly
  • custom → use visual notes verbatim (freeform)
• notes (optional) — extra visual details (outfit, palette, seed). For kind: custom, this becomes the character description.

Workflow

1. Compose the base character prompt

Pick the flavor template for kind, then concatenate and :