echogarden/docs/CUDA.md

An easy-to-use speech toolset. Includes tools for synthesis, recognition, alignment, speech translation, language detection, source separation and more.

# How to enable the `cuda` ONNX execution provider

On Linux, or the Windows Subsystem for Linux (WSL), you can enable CUDA-based GPU acceleration for a number of engines, by selecting the `cuda` ONNX execution provider, usually thorough an option like `[engine].provider=cuda`.

Engines where CUDA is supported include:
* `whisper` (speech recognition)
* `kokoro` (speech synthesis)
* `vits` (speech synthesis)
* `silero` (speech recognition)
* `mdx-net` (source separation)

## Steps
* Download and globally install [CUDA Toolkit 12.x](https://developer.nvidia.com/cuda-downloads) based on the instructions in the linked download page
* Download and globally install [cuDNN 9.x](https://developer.nvidia.com/cudnn-downloads) based on the instructions in the linked download page
* Reinstall Echogarden

Reinstallation is needed because the `onnxruntime-node` dependency package runs a post-install script that attempts to detect the global installation of CUDA, and then conditionally downloads additional files required for it to work.