avatar-video

# Talking-Avatar Video (Aurora) ## Purpose Aurora turns **(an avatar OR a face image) + a spoken script (+ an optional voice)** into a talking-avatar video. This is the tool for "make an avatar say this", spokesperson clips, UGC-style talking heads, and explainer presenters. This is NOT the tool for: - general b-roll / cinematic / product clips → use `asset-generation`; - building a reusable avatar identity → use `avatar-create`. ## Inputs - **A subject**: an existing `avatar_id` (from `avatar_list`, or one you built via `avatar-create`) OR a face image. - **A script**: the text to speak. - **A voice** (optional): pick one from `voice_list`, or a `clone_id` from `voice-clone`. Voices listed by `voice_list` are the ones usable here. Read the `aurora_generate` schema for the exact parameters (aspect ratio, subject/image field, voice field, etc.). Do not guess params. ## Flow 1. Resolve the subject: choose an avatar (`avatar_list`) or confirm the face image. If the user wants a brand-new avatar identity first, go through `avatar-create`, wait until it appears in `avatar_list`, then return here. 2. Resolve the voice with `voice_list` (or clone one via `voice-clone`). 3. Call `aurora_generate` with the subject + script + voice. 4. Poll `aurora_get` until it reports done (or failed). 5. Surface the finished video URL. ## Align Before Generating Aurora renders are paid and take time. Before generating, confirm the load-bearing choices: the exact script text, which avatar, which voice, and the aspect ratio / format. If any is a real fork (e.g. voice choice, vertical vs landscape), align first rather than guessing. ## Async & Verify `aurora_generate` starts a render; it is not ready synchronously. Poll `aurora_get` and only report success once it returns a finished video URL. On failure, report the reason. See `verification` and `known-errors`.