HeyGen text to speech API: endpoint, limits and price

Yes: HeyGen's API has a text to speech endpoint that returns audio without making a video. POST /v3/voices/speech takes 1–5,000 characters and a voice_id from HeyGen's Starfish-compatible voices, and returns an audio_url, the audio's duration and optional word timestamps. HeyGen's API pricing lists this Starfish speech at $0.12 per minute. Professional voice clones use a separate HeyGen Voice endpoint.

Everything about HeyGen below comes from its developer docs and its API pricing article, read on 2026-09-28 and linked in each caption. The Sume section at the end comes from the TTS 1.0 schema in the Sume API reference.

What does a HeyGen text to speech request take?

Send the request with your HeyGen API key in the X-Api-Key header. A successful call answers 200 with the audio_url in the response body.

How do HeyGen's professional voice clones speak?

A professional clone runs on HeyGen Voice, not Starfish, so it has its own endpoints and terms, per the HeyGen Voice and HeyGen Voice Speech pages:

• It is a paid feature: each professional voice takes a purchased voice clone slot, trained from 1–10 recordings of one speaker totaling at least 20 minutes.
• POST /v3/models/audio/tts waits and returns one mono 44.1 kHz WAV; POST /v3/models/audio/tts/stream streams ordered audio parts as Server-Sent Events.
• text is 1–5,000 characters and language is required. Each endpoint allows 30 requests per minute per workspace member.
• An instant clone, made from a single recording in minutes, runs on Starfish and works with POST /v3/voices/speech like any catalog voice.

How much does HeyGen text to speech cost?

HeyGen's API pricing article lists Speech on Starfish at $0.12 per minute (2 credits). HeyGen Voice synthesis costs 0.6 API credits per generated minute, on top of the voice slot. How HeyGen's pay-as-you-go API credits work is in HeyGen API pricing.

Can I use the speech in a HeyGen avatar video?

Yes. HeyGen's Audio to Video guide says the audio_url from POST /v3/voices/speech can be passed straight into audio_url on POST /v3/videos, or you can skip the audio step and send script plus voice_id. The separate TTS call helps when you reuse one narration across several videos. Audio to avatar AI covers the general pattern.

What if I only need speech from a TTS API?

If you only need audio, a general TTS API also works. Sume's TTS 1.0 is one option, with different limits:

• POST /v1/tts-1.0/generate takes a transcript of up to 20,000 characters, with the voice set by voice.id or by an avatar's voice through avatar_id or avatar_handle.
• Set language for every non-English transcript; omitted, it defaults to English. timestamps.words: true returns word timings.
• It runs as an asynchronous job with polling or a webhook, not a stream.
• It costs $0.0475 per 1,000 characters, spaces and punctuation included, plus a 5.5% agent fee by default. Text to speech API has the full request.