audio.cpp
An all-in-one, pure C++ inference engine for audio models, powered by ggml. Supports TTS, STT, VAD, voice conversion, music generation, and more, with highly op
What is audio.cpp?
An all-in-one, pure C++ inference engine for audio models, powered by ggml. Supports TTS, STT, VAD, voice conversion, music generation, and more, with highly optimized performance. No Python dependency.
Overview of audio.cpp
What it is
audio.cpp is a high-performance C++ audio inference framework built on top of ggml. It is designed to make modern local audio models practical, portable, and fast. The framework supports Windows, Linux, and macOS, with backends for NVIDIA, AMD, Apple Silicon, and CPU-only machines. It provides a shared native runtime for over 100 audio model families and 190+ variants, covering TTS, ASR, VAD, voice conversion, music generation, and more. No Python dependency is required.
Key features
- Parity: Strong parity tooling against Python reference paths.
- Performance: Performance-focused execution, reusable sessions, and batch-style offline inference. Optimized for CUDA.
- Portability: A portable native stack centered on ggml, with CUDA, HIP/ROCm, Vulkan, Metal, and CPU backends behind shared CLI and server entry points.
- Pipelines: Experimental JSON pipeline support for higher-level multi-step workflows.
- Audio Utilities: Built-in denoise, enhancement, resampling, and STFT/ISTFT utilities for production-style task paths.
Use cases
audio.cpp is used for text-to-speech (TTS), automatic speech recognition (ASR), voice activity detection (VAD), voice conversion, music generation, source separation, forced alignment, speaker diarization, audio codec tasks, and higher-level workflows such as speech-to-speech conversations. It also supports production deployment examples like Fun-ASR-Nano and SenseVoice on the FunASR platform.
Getting started
See the repository README for installation and quick-start instructions.
License and pricing
The license is not stated (NOASSERTION). The project is open source.
Based on the project's public README and repository metadata.