Summary

Add a first lightweight persona pack surface so users can personalize OpenHuman as one coherent identity: persona text, mascot avatar, and voice selection.

Motivation

OpenHuman already has most of the ingredients separately:

• workspace prompt files such as SOUL.md for personality and tone
• user-selected ElevenLabs voice IDs
• mascot rendering with state/viseme support and backend mascot manifests

What is missing is a simple user-facing bundle that makes these feel like one setup flow.

Persona Pack v1 scope

• Let users configure a persona display name / description in the product surface.
• Let users edit or reset SOUL.md from the app.
• Let users keep using an existing ElevenLabs voice ID for speech.
• Let users set a custom mascot avatar from a GIF URL or local file path.
• Treat GIF avatars as a simple fallback renderer: animated image only, no lip-sync or state-specific visemes.
• Persist the selected persona/avatar/voice preferences locally.

Follow-up scope

• Optional state-specific avatar assets: idle, thinking, listening, speaking, sleeping.
• Import/export folder manifest for a full persona pack.
• Provider-backed voice clone flow with explicit consent, upload validation, deletion controls, and status UI.
• Local voice cloning research if a safe on-device model becomes viable.

Non-goals for v1

• No voice cloning from uploaded audio files.
• No remote hotlinking without validation/caching strategy.
• No guarantee that arbitrary GIFs support lip-sync or speaking/listening states.

Acceptance criteria

• Users can provide a GIF URL/local path as a mascot override.
• Invalid or unsafe avatar values are rejected or reset to default.
• Existing mascot color, backend mascot, and voice ID preferences continue to work.
• PR includes focused tests for persistence/validation/render selection.