Rave Training
RAVE is a tool that allows for real-time audio generation. This is guidance on how to setup a Python environment that is suitable for training a RAVE model.
A Linux machine is required for training. However, once your training is complete, it should be possible to export a model that may be used to generate audio in MacOS, Raspberry Pi, Windows and Linux. The model will have to be loaded in Max/MSP or Pure Data with the nn~ extension or SuperCollider with the nn.ar extension.
RAVE on a Raspberry Pi
Be aware that you must export your model with a particular argument if you wish to use it on a Raspberry Pi.
Platforms
- ❌ Windows
- ❌ macOS
- ❌ Raspberry Pi
- ✅ Linux
Requirements
- A CUDA-compatible GPU. This cannot be a GPU that uses the newer Blackwell architecture.
- A collection of audio files that are alltogether at least an hour long. Go here to see IRCAM’s guidance on data preparation.
- libsox-dev