transcribe.cpp
C/C++ library for running many speech recognition models on ggml
What it is
transcribe.cpp is a C/C++ inference library that runs GGUF speech-to-text models such as Whisper, Parakeet, Canary, Moonshine, Qwen3-ASR and Voxtral on the ggml runtime. It supports streaming and batch transcription, plus side models for speaker diarization, language identification and voice activity detection, with Metal, Vulkan, CUDA and CPU backends. Developers building local transcription into desktop apps or services can pick from dozens of verified, quantized variants published on Hugging Face.