Voices

Voice selectors

Selector	Result
`default`	Uses the server default voice
`voice_...`	Uses a public catalog voice
Reference audio	Zero-shot clones the speaker for one request

List public voices

voices = client.voices.list()

Select a returned ID and pass it to the speech request.

voice = next(item for item in voices.data if item.name.casefold() == "ira")

speech = client.audio.speech.create(
    model="teen-v1",
    voice=voice.id,
    input="Use a public voice.",
    response_format="mp3",
)

Voice IDs are opaque. Store the returned ID instead of building one from a display name.