audio.cpp

audio.cpp is an all-in-one, pure C++ inference engine for audio models, powered by ggml. It eliminates the need for complex Python environments, providing a shared native runtime for various audio tasks. Key features include: - **Broad Model Support**: Supports over 80 model families and 120+ variants for Text-to-Speech (TTS), Automatic Speech Recognition (ASR), Voice Activity Detection (VAD), voice cloning, voice conversion, music generation, source separation, and speaker diarization. - **High Performance**: Optimized for CUDA, Metal, ROCm, Vulkan, and CPU. It achieves significant speedups (up to 8x faster than Python paths) and reduced latency. - **GGUF Format**: Fully supports the GGUF model format, with optimized Q8 and other quantized packages for efficient memory usage and speed. - **Portability**: Runs on Windows, Linux, and macOS, supporting NVIDIA, AMD, and Apple Silicon GPUs. - **Production Readiness**: Features reusable sessions, batch offline inference, streaming support, built-in audio utilities (denoise, resampling, STFT), and JSON pipeline support for multi-step workflows. - **User Interfaces**: Includes a command-line interface (CLI), an API server, and a native WebUI with an Arena tab for side-by-side model comparison.