Transcript unusable
transcript_unusable
The audio decoded, but the engine refused to keep the result, for one of two reasons (the detail says which). Either speech recognition got stuck repeating the same words over most of the recording, even after one re-decode of the looping part; or no speech was detected at all because the clip is silent or holds only a tone, noise or a music bed. VoiceLabs refuses both rather than returning text that was never spoken, saves no capture for it, and refunds the unit.
Why it happens
Looping: a known speech-recognition failure on long stretches of silence, near-silence or heavy background noise (music, hum, a crowd). No speech: the recording itself has nobody speaking in it, often a microphone that was not capturing. Both are properties of the audio, not of your request body or your credential.
How to fix it
Looping: send the same audio once more, which can come back clean; if it fails again, trim the silent or noisy parts, or send a cleaner recording. No speech: record again with someone speaking; the same audio will fail the same way.
Safe to retry?
Looping: once, yes, because the re-decode samples and a second attempt can succeed. No speech: no, retrying the same audio cannot help; send a recording with speech. A loop of retries on the same audio helps in neither case; change the audio instead.
What it looks like on the wire
{
"type": "https://voicelabs.now/errors/transcript_unusable",
"title": "Transcript unusable",
"status": 422,
"detail": "A human-readable explanation of this particular occurrence.",
"instance": "/v1/speech",
"code": "transcript_unusable"
}
Branch on code, never on status or on the prose in detail: two different problems can share a status, and detail is written for humans and may be reworded. Problems a settings page can resolve also carry settings_url, and a scope failure carries required_scope.
Every VoiceLabs API error is an RFC 9457 problem document with a stable code.