Docs
Transcription
Turn audio into text. Most often used for Korean meeting minutes and support call records.
Calling
POST /audio/transcriptions with the file as multipart. Specifying language improves accuracy.
| Field | Type | Note |
|---|---|---|
| file | binary | multipart, max 25MB |
| language | string | ISO code |
| timestamp_granularities | array | word | segment |
| diarize | boolean | Split by speaker |
Timestamps
Use timestamp_granularities for word or sentence level timing. Useful when building subtitle files.
Speaker separation
Enable diarize to split segments by speaker. Only supporting models become candidates.