vibevoice.cpp: Microsoft VibeVoice (TTS + long-form ASR with diarization) ported to ggml/C++, runs on CPU/CUDA/Metal/Vulkan, no Python at inference
A few weeks ago I shipped vibevoice.cpp, a pure-C++ ggml port of Microsoft VibeVoice (the speech-to-speech model with voice cloning, https://github.com/microsoft/VibeVoice). Wanted to post a follow-up here because we're at a point where the engine …