whisper.cpp is an open-source port of OpenAI's Whisper model to plain C/C++. Whisper is an automatic speech recognition (ASR) model: a neural network that turns spoken audio into written text. The project belongs to the ggml-org organization on GitHub. The Whisper code itself sits in two files; the rest is the ggml machine learning library.
The program runs on the device itself, without dependencies, on macOS, iOS, Android, Linux, Windows, WebAssembly and Raspberry Pi. A developer downloads a Whisper model converted to the ggml format, builds the whisper-cli example and passes it a 16-bit WAV file to get the transcript. On Apple Silicon the inference runs on the GPU through Metal, and the encoder can run on the Apple Neural Engine through Core ML. Integer quantization, which stores the model weights with fewer bits, reduces memory and disk use. A C-style API lets other applications embed the library.
whisper.cpp is free and open source under the MIT license.