HeyGen generates video of AI avatars that speak a script. A builder supplies text, picks an avatar and voice, and receives a rendered video; the same platform drives interactive avatars that hold live conversations over WebRTC. Avatars come from a set of stock characters, from a photo, or from a digital twin trained on a person's own footage, and the Avatar V model keeps one character's face and expressions consistent across clips of different lengths.
The API is REST-based with webhook delivery for asynchronous renders, plus a CLI and MCP servers so coding agents can call it directly. Batch requests accept up to 100 items per call, and an enterprise tier adds higher concurrency and reserved compute.
Features
- Avatar video generation: turns a script into a spoken avatar video using avatar models III, IV, and V
- Interactive avatars: real-time streaming avatars for conversational agents over WebRTC
- Text-to-speech: the Starfish voice engine supplies voices for generated video
- Photo and image avatars: builds an avatar from a photo or converts an image to video
- Video translation: dubs videos into more than 175 languages with lip-sync
- Lip-sync replacement: replaces a video's audio track and re-syncs the mouth
- Templates and composition: template-based and studio-style scene assembly, with AI clipping, music, and sound effects
- Async delivery: webhooks report render completion; batches take up to 100 requests
