Send existing call recordings to speech recognition for transcription.

When to use. Transcribe calls that were recorded earlier but not yet processed. To retrieve a finished transcription for a single call, use GET /speech-recognition/{call_id}. To search transcribed calls and have results delivered to your own endpoint, use GET /speech-recognition with a callback URL.

Request body. Provide exactly one of call_ids, campaign_lead_ids or lead_ids (each an array) to select the calls, plus:

• token (string, required): your speech recognition token. To obtain a token, contact your account manager.
• languages (array, required): up to three language codes to transcribe.
• recording_side (string, required): which channel to transcribe. One of lead, agent, both.
• min_duration (integer): minimum recording length in seconds. Default 1.
• from, till (date-time): restrict to calls created within this range.

Returns. A nested results tree keyed by the selector you sent (calls, campaign_leads or leads). Each call entry reports whether it was accepted for transcription, or the reason it was skipped (for example the recording is still being prepared, or the call was already sent).

Note. Only the token used to submit a call can later retrieve that call's transcription.