The AI avatar API for products that generate video

Two endpoints and a webhook. Your product creates an avatar from a short video of a real person, then asks for as many videos of them as it needs. No studio, no seats, no watermark.

POST /avatars

Send a clip URL or an upload. We normalize it, clone the voice and return an avatar id.

POST /generations

Send text, a script with pauses, or your own audio. Get a 202 and a generation id back.

Signed webhook

The finished MP4, the audio and word-level timings arrive on your endpoint, HMAC-signed.

Built for code paths, not editors

Avatar studios are made for a person sitting in an editor. B2B tools need the opposite: a request per lead, per candidate or per event, and a file at the end. PitchTwin is only the API, so everything about it is designed around that loop.

  • Async by design: requests return 202 immediately, results arrive by webhook or polling
  • Consistent errors: every failure is { error: { code, message } } with a stable code
  • Every request is visible in the dashboard logs with tokens charged
  • Each avatar is pinned to its engine version, so faces never drift after an upgrade

What you can send

Plain text is voiced in the avatar's cloned voice. A script lets you add pauses where the speaker should breathe. Audio mode takes your own finished speech, for example from your own ElevenLabs account, and the avatar speaks it as is.

  • Text or script: up to 5,000 characters, up to 50 pauses of up to 10 seconds
  • Audio: up to 10 minutes, with optional character timings (ElevenLabs format accepted)
  • Voice options: speed 0.7–1.2 and an ISO 639-1 language code per request

Priced like infrastructure

You pay per second of finished video from prepaid tokens that never expire: from $0.74 a minute with your own audio and from $1.12 when we voice the script. Failed renders are refunded automatically.

Questions about ai avatar api.

How long does it take to integrate?

Most teams need two server functions and one webhook route: create an avatar, request a video, receive the result. The dashboard's Copy prompt button gives your coding agent everything it needs to wire it in.

Is there an SDK?

The API is plain REST with JSON, so any HTTP client works. The docs include a complete Node.js example and cURL for every endpoint.

What does the webhook contain?

A generation.succeeded or generation.failed event with the full generation object: video_url, audio_url, word timings, duration and tokens charged, signed with your webhook secret.