whisper.cpp

A plain C/C++ port of OpenAI's Whisper speech-to-text model with no Python and no dependencies, running fast enough on a laptop CPU to transcribe faster than real time. It runs on a Raspberry Pi and in a browser via WebAssembly. The practical value is that audio never leaves the machine, which makes it usable for interviews, medical notes and anything else you cannot upload. MIT, from the ggml author, and it feeds a large ecosystem of tools. Tradeoff: accuracy is bounded by the Whisper model itself, so heavily accented speech and noisy recordings still degrade, and larger models trade speed for quality in the way you would expect.