voice-clone

# Voice Clone & Voices ## Purpose Clone a voice from an audio sample, and browse the voices available to the workspace. Both cloned and catalog voices are consumed by `aurora_generate` (see `avatar-video`). ## Tools - **`voice_list`** — browse voices usable with Aurora. Returns id, name, gender, preview_url, is_cloned. Use it to pick an existing voice before cloning. - **`voice_clone`** — clone a voice from an audio sample. Returns `{clone_id, status}`. Read the schema for the audio-sample input. - **`voice_clone_get`** — poll a clone job; the voice/accent id populates once `done`. - **`voice_clone_widget`** — opens a record/upload panel; host-specific and does not apply to Codex. In Codex, take the audio sample through the normal upload path (see `media-upload`) or a provided URL, then call `voice_clone`. ## Flow 1. If the user just wants an existing voice, `voice_list` and pick one — no clone needed. 2. To clone: obtain the audio sample, call `voice_clone`, then poll `voice_clone_get` until done. 3. Use the resulting voice (or `clone_id`) in `avatar-video` / `aurora_generate`. ## Notes - Cloning consumes a workspace voice slot (in-app quota gates). If the slot is full or a creation-time cap is hit, surface that to the user rather than retrying blindly (see `known-errors`). - Don't clone without an explicit request — it's a stateful, quota-bound action. ## Async & Verify Poll `voice_clone_get`; only report success once the clone is `done` with a usable voice. See `verification`.