Whisper opens multilingual speech recognition
OpenAI releases speech-recognition models and inference code.
A historical account of this event. Usage notes are YLEM editorial analysis; check the provider for current features and prices.
What changed
Whisper supports multilingual transcription and speech translation into English. OpenAI released weights and inference code; this was a model release, not a ready-made meeting-notes application.
Sources: [1] OpenAI · Introducing Whisper
What this means in practice
Transcribe a short recording, correct names and terminology, then use the transcript to draft notes or subtitles. Keep timestamps so uncertain phrases can be checked against the recording.
How to read the result
A readable transcript does not prove accuracy. Check numbers, negations and speaker attribution; decide separately whether a recording may be uploaded to a third-party service.