When testing ElevenLabs, follow the response contract as carefully as the request. The text-to-speech conversion endpoint returns audio; a client that unconditionally parses successful responses as JSON will mishandle a valid generation.
Choose a supported output format and pass the returned bytes to a compatible playback or storage path. Use a short, non-sensitive sentence to test the full route from request to audible output.
Keep error handling separate from successful audio handling. A saved file with the expected extension does not prove that its contents are playable audio.
Akousa