Transcribe a recording
One recording in, its text out, in the same request. This is POST /v1/audio/transcriptions under its v2 name: the same route, so the OpenAI SDK works against https://sandbox.voice.miraiminds.co/v1 with only the key changed.
Up to 25 MB and 2 minutes of audio: WAV, MP3 and most audio formats; headerless audio needs encoding (linear16, mulaw, alaw) and sample_rate (8000 or 16000). A longer recording is a job for Transcriptions (POST /v2/transcriptions).
Billed per second of audio, rounded up, from your wallet; the answer’s X-Cost-Paise says what it cost and X-Request-Id is its row in List requests.
Authorizations
Your workspace secret key, sk_live_ followed by 32 hex characters. Create one in the console under Developers.
Body
The recording. One file per request.
mira-stt.
"mira-stt"
hi-IN, en-IN, gu-IN, mr-IN, pa-IN, bn-IN, ta-IN, te-IN, kn-IN or ml-IN (default hi-IN). The two-letter code works too.
"hi-IN"
json (default, {"text"}), verbose_json (with each utterance's start and end) or text.
Response
The text