audio_too_long · 3006 · HTTP 422

{
  "ok": false,
  "request_id": "req_…",
  "error": {
    "code": "audio_too_long",
    "number": 3006,
    "message": "The media is longer than the AI-transcription length limit, so no audio was downloaded and nothing was charged. This is about the media, never the account: a balance that cannot cover the media answers insufficient_credits instead. Distinct from no_speech, which means audio WAS transcribed and carried no speech. Retrying the same media changes nothing; the response's message names the limit.",
    "docs": "https://transcriptfetch.com/docs/errors/audio_too_long"
  }
}

Every failure carries error.code, error.number, error.message and error.docs, plus at most one of retry_with (the request change that would succeed), details (structured specifics) or issues (validation problems).

Too long to transcribe. The media is longer than the AI-transcription length limit, so no audio was downloaded and nothing was charged. This is about the media, never the account: a balance that cannot cover the media answers insufficient_credits instead. Distinct from no_speech, which means audio WAS transcribed and carried no speech. Retrying the same media changes nothing; the response's message names the limit.

Shown to end users as: “This video is longer than we can transcribe. Try a shorter video, or fetch its captions instead.”

• Too long to transcribe. The media is longer than the AI-transcription length limit, so no audio was downloaded and nothing was charged. This is about the media, never the account: a balance that cannot cover the media answers insufficient_credits instead. Distinct from no_speech, which means audio WAS transcribed and carried no speech. Retrying the same media changes nothing; the response's message names the limit.

• Permanent for this input: the same request will fail the same way, so do not retry unchanged. Failed requests are never charged.
• AI Fallback Transcription would not help here: the content itself is not accessible to fetch.