Get Voice Preview

Fetch ready-to-play preview audio for one voice in one language.

Returns the voice’s name, the preview’s locale, the model it was synthesized with, its stored content_type, and a sample_url — a time-limited presigned URL valid for 1 hour; regenerate it by calling this endpoint again rather than caching it long-term. Pass model to prefer a specific TTS model; a preview from another model is still returned when that model has none.

language accepts a bare family (ko, en) as well as an exact locale (ko-kr, en-gb). A bare family expands to every supported locale in it, and which region wins is deterministic but arbitrary — so the response echoes the locale actually served. Read locale, never the request parameter, when labelling what the caller is hearing.

404 means no preview audio has been generated for this voice in that locale — NOT that the voice cannot speak it. supported_languages on the voice is the claim about what it can speak; preview_locales is the list of locales this endpoint will succeed for. 404 is also returned when the voice does not exist, is not accessible to the caller’s workspace, or is a platform voice from a provider that is not available to your account.