Audio
OpenAI-compatible transcription. The upload is bounded at 25 MiB and must be audio/wav, audio/mpeg, audio/flac, or audio/ogg. The request waits for a terminal Job and returns the transcript projected into the requested response_format; past the wait deadline it returns 202 with the job_id and the client continues through the standard Job and Artifact APIs. X-Job-Id is always set once admission succeeds. Failures use the OpenAI-compatible error envelope, not the platform ErrorEnvelope. prompt and temperature are accepted, bounded, and ignored, and stay out of the idempotency digest. project_id is optional only where server-side default Project resolution is wired and the principal's Membership is unambiguous; a caller in several Organizations must send it (project_ambiguous, project_required, default_project_not_found).
Authorization
DeveloperCredential In: header
Header Parameters
length <= 128Request Body
multipart/form-data
audio/wav, audio/mpeg, audio/flac, or audio/ogg, at most 25 MiB.
binary1 <= length <= 128length <= 128length <= 32"json"Value in
- "json"
- "text"
- "verbose_json"
- "srt"
- "vtt"
Selects the billing class the resolved model must match. A mismatch answers 404 model_not_found.
"accurate"Value in
- "fast"
- "accurate"
"segment"Value in
- "segment"
- "word"
Extension field. Omitted, the façade resolves the principal's default Project where that is unambiguous.
length <= 128Optional extension. The server measures the media itself when omitted, and verifies any supplied value against the received bytes so a client cannot understate audio-seconds.
1 <= value <= 14400Accepted, bounded, and ignored. Excluded from the idempotency digest.
Accepted, bounded, and ignored. Excluded from the idempotency digest.
0 <= value <= 1Streaming is not supported here; any other value answers 400 unsupported_parameter. Poll the returned job instead.
Value in
- "false"
Response Body
application/json
application/json
application/json
application/json
application/json
application/json
application/json
application/json
application/json
curl -X POST "https://example.com/v1/audio/transcriptions" \
-F file="string" \
-F model="string"{
"text": "string",
"task": "transcribe",
"language": "string",
"duration": 0,
"segments": [
{
"id": 0,
"start": 0,
"end": 0,
"text": "string"
}
]
}