This repository contains code used to implement a standalone desktop application with a web UI that embeds an existing machine learning model for timbre transfer. It also uses interactive machine learning to train custom gestural control for sound generation, a method popularised by Rebecca Fiebrink through Wekinator.
This workshop aims to explain the steps leading to the creation of Style Transfer Demo App.
The workshop materials are based on a collaborative effort by Dynamic Cast in preparation for a workshop at Audio Developer Conference in 2024.
This app loads a RAVE model and enables navigating its latent space with controls.
- Download one of the RAVE models from here. This app is tested with Darbouka_onnx
- Set the path to the model in
instance/application.cfg - Install the dependencies:
pip install -r requirements.txt - Run the app:
flask run
In order to run the app in debug mode, run:
flask run --debug