# Get a voice sample

`GET https://api.eesi.ai/v1/voices/{voice_id}/sample`

A short-lived signed URL for the voice's reference audio.

The reference clip *is* an authentic sample of the voice (for a built-in it
is the frozen design output; for a clone it is the user's own upload), so
playback costs a storage read rather than a GPU synthesis. Returning the URL
rather than the bytes keeps audio off the API process and lets the browser
feed it straight to an ``<audio>`` element — the same pattern call
recordings use. ``POST /{id}/preview`` still exists for hearing the voice
read a fresh sentence.

Parameters:

- `voice_id` (path, required)

Responses: 200, 400, 404.

Schemas: https://docs.eesi.ai/api-reference/public.json
