Duyu

Output formats

Same call, different response_format: plain text for dictation, word timestamps for archives, subtitles for video.

response_format	Content	Content-Type
json	`text` only.	application/json
text	Plain text with a trailing newline.	text/plain
verbose_json	`text`, `language`, `duration`, `segments[]`, and `words[]` on request.	application/json
srt	SubRip subtitles.	text/plain
vtt	WebVTT subtitles.	text/vtt

verbose_json

curl https://voice.patientdesk.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer pd_live_..." \
  -F file=@toplanti.mp3 \
  -F response_format=verbose_json \
  -F "timestamp_granularities[]=word"

{
  "task": "transcribe", "language": "tr", "duration": 512.4,
  "text": "Merhaba, ben Ayşe Yılmaz. Sipariş numaram CNB123...",
  "segments": [{ "id": 0, "start": 0.0, "end": 4.2, "text": "Merhaba, ben Ayşe Yılmaz." }, ...],
  "words": [{ "word": "Merhaba,", "start": 0.0, "end": 0.6, "probability": 0.98 }, ...]
}

segments[]object[]

id, start, end, text. Sentence-level timing.

words[]object[]

word, start, end, probability. With timestamp_granularities[]=word.

durationnumber

Audio length in seconds. This is the billed duration.

Subtitles

srt

1
00:00:00,420 --> 00:00:03,100
Yarın saat on dörtteki randevumu on altıya alabilir miyiz?

2
00:00:03,600 --> 00:00:05,020
Tabii, hemen bakıyorum.

Subtitle cues follow segment boundaries and are wrapped to at most 42 characters and 2 lines.