whisper.cpp

whisper.cpp runs OpenAI's Whisper speech recognition models in plain C and C++. It takes an audio file or microphone input and produces a transcript, with timestamps, without sending audio anywhere. Smaller model sizes run on a laptop CPU, and larger ones benefit from a GPU or Apple silicon.

Many transcription apps for Mac and Windows are built on it, and developers embed it in their own tools. It can output subtitle files, handle many languages and translate speech into English text. Accuracy depends on audio quality, accents and the model size you pick.

It suits people who need private transcription of interviews, meetings or voice notes and are happy with a command line, or who want to build their own app. Read transcripts before relying on them for quotes or records, since speech models can mishear names and numbers.