HeyGen Instant Clone

Clone a speaker’s voice from a single recording in minutes on HeyGen’s Starfish engine, then use it for text-to-speech and avatar videos.

Full schema: POST /v3/voices/clone.

An instant clone creates a usable voice from one recording in minutes. It runs on the Starfish engine, so the voice works wherever Starfish-compatible voices do: Third Party Speech and POST /v3/videos. For a studio-grade clone trained on the HeyGen Voice model, see HeyGen Professional Clone. Instant cloning on third-party voice engines is coming soon.

Create the clone

curl -X POST "https://api.heygen.com/v3/voices/clone" \
  -H "X-Api-Key: $HEYGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "voice_name": "Customer narrator",
    "audio": {
      "type": "url",
      "url": "https://example.com/recording.wav"
    }
  }'

Response

{
  "data": {
    "voice_clone_id": "1bd001e7e50f421d891986aad5c8bbd2"
  }
}

audio is a public HTTPS URL, a HeyGen asset_id from the Assets API, or inline base64 with a media_type. Optionally pass a language hint such as es (auto-detected when omitted); remove_background_noise defaults to true. Each account has a clone allowance, and a request past it returns 400 resource_limit_reached; deleting a clone with DELETE /v3/voices/{voice_id} frees the slot.

Poll clone status

Cloning runs asynchronously. Poll GET /v3/voices/{voice_clone_id} until status is complete:

curl "https://api.heygen.com/v3/voices/1bd001e7e50f421d891986aad5c8bbd2" \
  -H "X-Api-Key: $HEYGEN_API_KEY"

Status	Meaning
processing	The clone is still being created.
complete	The voice is ready to use.
failed	Cloning ended unsuccessfully. Retry with a cleaner recording.

Use the voice

Once complete, the voice_clone_id works anywhere a voice_id does: pass it to Third Party Speech for standalone audio, or to POST /v3/videos to narrate an avatar video. Cloned voices appear in GET /v3/voices?type=private.