Whisper is OpenAI's AI model for converting spoken words into written text. It works in more than 99 languages and approaches human-level accuracy. Released as open-source software in September 2022, it is widely used for transcribing meetings, interviews, podcasts, and videos.

Whisper was trained on 680,000 hours of multilingual audio collected from the web. It is available in multiple sizes — from a "tiny" version suited to devices with limited computing power to a "large" model for maximum accuracy. Whisper Large V3 (released November 2023) was trained on over 1 million hours of audio and reduced errors by 10–20% compared to prior versions. OpenAI made the model and code available under a permissive open-source license; it is also accessible through the OpenAI API. In March 2025, OpenAI released newer transcription models based on GPT-4o with even lower error rates.