API reference · Voice Intelligence
Transcription
Turn a short recording into text. Common uses are transcribing a voicemail reply, checking what a recorded message actually says before a campaign, or indexing call audio.
POST /voice/public/asr/transcribe · scope media:read
curl https://api-v2.dropcowboy.com/voice/public/asr/transcribe \
-H "x-key: $DROPCOWBOY_KEY" -H "x-secret: $DROPCOWBOY_SECRET" \
-H "Content-Type: application/json" \
-d '{ "media_id": "1b7e3c9a-4d2f-4a8b-9e6c-7f2a1d5b3c80" }'
{
"data": {
"text": "Hi, this is Jordan returning your call about the appointment.",
"language": "en"
},
"meta": { "request_id": "c2a6e8f4-3b9d-4c1e-8a7f-5d3b1e9c6a24" }
}
| Field | Notes |
|---|---|
media_id |
An MP3 or WAV in your media library. |
ext |
The extension the media_id file was uploaded with: .wav (default) or .mp3. |
url |
Public URL of a WAV file (PCM 16-bit). sample_url is accepted as an alias. |
Send media_id or url.
Audio formats
media_idtakes any MP3 or WAV you uploaded as media. A few seconds after the upload completes we make a normalized copy (8 kHz mono WAV), and that copy is what gets transcribed. Until it exists, an uploaded WAV is transcribed as is and an uploaded MP3 answers409. Wait two or three seconds and retry.urlmust point at a WAV file with PCM 16-bit samples. Any other format answers400. For MP3 or other formats, upload the file as media and passmedia_idinstead.
Audio can be up to 10 MB. The call is synchronous, so keep recordings to a few minutes. The request is cut off at the gateway's 29-second limit.
Errors
| Status | Cause |
|---|---|
400 |
Neither url nor media_id was given, the url audio is not WAV, the audio is larger than 10 MB, or it could not be fetched or transcribed. |
402 |
No funds for speech to text (payment-required). |
404 |
The media_id does not exist or its file was never uploaded. |
409 |
The media_id file is an MP3 still being converted. Retry in a few seconds. |
502 |
Speech to text is temporarily unavailable. Safe to retry. |
See Errors and limits.