HeyGen Voice

HeyGen Voice is HeyGen’s in-house voice model, powering professional voice clones and direct completed or streaming speech synthesis.

The HeyGen Voice APIs are in private preview. Your account must be enabled before these endpoints are available — see how to request access.

HeyGen Voice is HeyGen’s in-house voice model. It trains a dedicated voice adapter per speaker and synthesizes speech directly from that adapter, giving the highest-fidelity voice output HeyGen offers.

What it does

• Professional voice cloning — train a dedicated voice from 1–10 recordings of the same speaker totaling at least 20 minutes. Training runs asynchronously; poll until the voice is ACTIVE. See HeyGen Professional Clone. For cloning from a single short recording, see HeyGen Instant Clone on the Starfish engine.
• Speech generation — synthesize speech from an ACTIVE voice as one completed 44.1 kHz WAV, or stream ordered audio parts over Server-Sent Events with optional word-level timestamps. See HeyGen Voice Speech.

Endpoints

Endpoint	Purpose
POST /v3/models/audio/voices	Create or retrain a professional voice.
GET /v3/models/audio/voices/{voice_id}	Read a voice and its training status.
GET /v3/models/audio/voices	List your HeyGen Voice audio voices.
DELETE /v3/models/audio/voices/{voice_id}	Delete a voice.
POST /v3/models/audio/tts	Generate one completed audio file.
POST /v3/models/audio/tts/stream	Stream audio parts as they are generated.

HeyGen Voice and third-party voices

The voices catalog (/v3/voices) covers HeyGen’s 300+ stock voices, designed voices, and HeyGen instant clones on the Starfish engine, synthesized through Third Party Speech. HeyGen Voice is HeyGen’s own model surface: its professional voices are created and used through the /v3/models/audio endpoints, scoped to the workspace that trained them.