Text-to-speech models
The speech model, the voices it speaks in, and the fields it accepts.
Chorus synthesizes speech in a preset voice, a voice cloned from reference
audio, or a voice designed from a written description. Its model id is
chorus.
| Voice | Request |
|---|---|
| Preset | voice set to a preset id on POST /v1/audio/speech |
| Cloned | ref_audio and ref_text inline, or voice set to the id of a voice created from a clip on POST /v1/audio/voices |
| Designed | voice set to the id of a voice created from a POST /v1/audio/voices/design preview |
chorus is the model on POST /v1/audio/speech, GET /v1/audio/voices,
POST /v1/audio/voices, and POST /v1/audio/voices/design.
Voices has the request shape for each voice.
Parameters
| Parameter | Notes |
|---|---|
input | The text to speak. Required. |
voice | A preset id, or the id of a voice your organization created. |
instructions | Style guidance on how the voice reads. Accepted for voices whose model supports it. |
response_format | mp3, opus, flac, wav, or pcm. |
speed | Speaking rate multiplier, 0.25 to 4.0. |
language | Output language. See supported languages. |
stream_format | sse or audio. Omit to receive the whole clip at once. |
ref_audio | Inline data: URL of reference audio, for cloning. |
ref_text | Transcript of the reference audio. |
temperature | Sampling temperature, 0 to 2. |
top_p | Nucleus sampling probability mass, 0 to 1. |
top_k | Number of top candidates sampled from, 0 or greater. |
timestamp_granularities | word, on the WebSocket. |
API has types, requiredness, and error codes. The model catalog carries the live list with per-model detail, and pricing has the current rates.