Guides 8 min read· 31 Jul 2026· By Guides
Speech-to-text with Whisper-class models
Practical transcription: chunking, diarization, and cost per minute.
Transcription is one of the highest ROI AI features you can add. Meeting notes, voice memos, podcast search — all unlocked by a single endpoint.
The call#
python
audio = open("meeting.mp3", "rb")
r = client.audio.transcriptions.create(model="whisper-large-v3", file=audio)Chunking long files#
Anything over 25MB should be split at silence boundaries. Overlap chunks by 2s to preserve word boundaries.
Cost: about ₹0.30 per minute of audio on Whisper large.