Speech-to-text transcription for voice messages and audio clips. Upload a wav, mp3, m4a, ogg or webm file as multipart form-data and get the spoken text back as JSON. Runs a dedicated ASR model, so short noisy voice notes transcribe accurately. Accepts clips up to 10 minutes / 25 MB, billed per 10-second block. Use it to turn voice notes, call recordings, podcast clips or voicemail into text an agent can read.