groq/whisper-large-v3
OpenAI's Whisper large-v3 speech-to-text, run on Groq: the most accurate Whisper model, robust to noise, accents and technical terms, in 99+ languages.
Takes flac, mp3, m4a, mpeg, mpga, ogg, wav or webm.
| Slug | groq/whisper-large-v3 |
| Kind | transcription |
| Vendor | groq |
| Endpoint | POST /v1/audio/transcriptions |
Charged per job from what the generator reports it rendered — your Logs show the exact figure for each one; see balance and billing.
Calling it
curl https://api.sociaro.com/v1/audio/transcriptions \
-H "Authorization: Bearer $SOCIARO_API_KEY" \
-F model=groq/whisper-large-v3 \
-F file=@./recording.mp3Parameters
| Field | Type | Required | Default | Values | What it does |
|---|---|---|---|---|---|
file | audio file (multipart) | yes | — | flac · mp3 · m4a · mpeg · mpga · ogg · wav · webm | The recording to transcribe, sent as a multipart form field. |
language | string | no | detected | an ISO-639-1 code, e.g. en, de, ru | The spoken language. Naming it improves accuracy and speed; leave it out to let the model detect it. |
prompt | string | no | — | text | Context or spelling to steer the transcript, such as names and terms, in the audio's language. |
response_format | string | no | json | json · verbose_json · text | `verbose_json` adds segments with timestamps. |
temperature | number | no | 0 | 0 – 1 | Higher is more varied; 0 is the most deterministic. |