handy-computer/transcribe.cpp
Summary
transcribe.cpp is an MIT-licensed C/C++ speech-to-text inference library that runs multiple model families via GGUF models on the ggml runtime, with GPU backends (Metal, Vulkan, CUDA) and a CPU path. It provides pre-built model support, bindings for multiple languages, and a structured project layout with documentation, tests, and examples. The project targets high-performance transcription across platforms and emphasizes verifiability of model quality (WER checks) and cross-language accessibility through bindings.