stable-audio-tools is an open-source PyTorch library that provides training and inference code for generative models used in conditional audio generation. It includes automatic encoders, implicit diffusion models, MusicGen, and more. Support for multi-GPU training allows generation of high-quality audio.