Overview

Whisper was first released by OpenAI in September 2022, trained on 680,000 hours of multilingual audio from the web. The large-v3 version, announced in November 2023, improved accuracy across many languages. Whisper is widely used in transcription apps and as a component in voice assistants.

Highlights

  • Speech recognition in many languages
  • Translation of speech into English text
  • Timestamps for subtitles
  • MIT license for commercial use
  • Many optimized community implementations

Typical uses

Transcribing meetings, interviews and podcasts, generating subtitles and building voice interfaces, including fully offline.

Things to consider

  • Can occasionally produce text that was not spoken, especially during silences.
  • Large model is slow on CPUs; smaller or optimized versions help.
  • Speaker identification requires additional tools.

Specifications are taken from the provider’s official announcement, documentation or model card. Benchmark results depend on evaluation settings; see how AI benchmarks work.