ID of the model to use. Only whisper-1
(which is powered by our open source
Whisper V2 model) is currently available.
Optional
promptAn optional text to guide the model's style or continue a previous audio segment. The prompt should be in English.
Optional
response_The format of the transcript output, in one of these options: json
, text
,
srt
, verbose_json
, or vtt
.
Optional
temperatureThe sampling temperature, between 0 and 1. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. If set to 0, the model will use log probability to automatically increase the temperature until certain thresholds are hit.
The audio file object (not file name) translate, in one of these formats: flac, mp3, mp4, mpeg, mpga, m4a, ogg, wav, or webm.