Transcription for legal work

Produce searchable draft transcripts of depositions, hearings, client interviews and recorded calls, with speakers separated and every word timed so you can cite and check the exact moment.

Get an API key Quickstart Supported languages

Last updated: · Markdown

A request for a multi-party recording

{
  "audio_url": "<fileUrl from POST /upload>",
  "model": "oracle_max",
  "include_diarization": true,
  "pro_diarization": true,
  "num_speakers": 4,
  "word_timestamps": true,
  "output_format": "json"
}

Use oracle_max with Pro diarization for rooms with several speakers and crosstalk. For recordings where each party has its own microphone channel, use channel_speaker_separation with channel_labels instead of diarization.

How teams use the output

  • Search hours of recordings for a name, date or phrase.
  • Cite a statement by its timestamp and jump to it in the audio.
  • Prepare summaries and exhibits from the draft, then verify against the recording.
  • Translate foreign-language recordings with translate_to on an Oracle model for a working understanding.

Limits

A machine transcript is a draft, not a certified court transcript, and translation output is not a sworn translation. Have a qualified person verify any passage you rely on. You are responsible for consent to record and for confidentiality obligations. KolWrite makes no certification claim for this API.

Questions and answers

Is the output a certified transcript?

No. It is a machine draft for search and review.

Can speakers be named instead of "Speaker 1"?

Diarization returns generic labels; map them to names after review. With one speaker per channel, channel_labels sets the names directly.

Can KolWrite transcribe phone recordings?

Yes. For stereo call recordings, channel separation labels each side.

Related pages