Duyu

Transcription

Send an audio file, get the text. Same contract as OpenAI audio/transcriptions; your existing Whisper client works unchanged.

POSThttps://voice.patientdesk.ai/v1/audio/transcriptions

Try it

curl https://voice.patientdesk.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer pd_live_..." \
  -F file=@kayit.m4a \
  -F response_format=json

Form fields

Send multipart/form-data. wav, mp3, m4a, webm, ogg and flac are accepted. Files longer than 30 s are split and re-joined automatically.

filefilerequired

The audio file. Up to 100 MB and 2 hours.

modelstring

duyu-1. whisper-1 is accepted for compatibility.

language"tr" | "auto"

Default tr.

promptstring

Spelling and vocabulary hint: drug names, person names, ID numbers. The model prefers these spellings.

response_formatstring

json (default), text, verbose_json, srt, vtt. See Output formats.

timestamp_granularities[]string[]

segment and/or word; only with verbose_json.

temperaturenumber

0–1. Default 0.

Response

{ "text": "Merhaba, ben Ayşe Yılmaz. Sipariş numaram CNB123." }

For short, single-speaker recordings, putting the expected terms in prompt noticeably reduces domain errors.