Living Looper

This repository is the Python CLI for the Living Looper, which can be used to export nn~ compatible Living Looper files.

For the instrument itself, see living-looper-sc

Python scripts

Using export.py, you can create your own Living Looper models from an nn~ compatible (causal) encoder-decoder model, e.g. RAVE.

• bring your own pretrained RAVE model
• clone this repo and create a Python environment with nn_tilde and pytorch
• run your RAVE model through export.py to produce a living looper .ts file

NIME paper

This project was presented at NIME 2023. The full paper is available here

Models

A Living Looper instance includes a pretrained RAVE encoder-decoder and a living loop algorithm, both implemented with pytorch and exported as a TorchScript file. Plugins for music software define a real-time audio and GUI interface, but must load a .ts file which defines the sound and behavior.

You can make your own living looper .ts models from any nn~ compatible streaming encoder/decoder model (such as these) using the export.py script.

Repository Structure

export.py - Python script which converts a RAVE model to a Living Looper model

living-looper-sc/ - SuperCollider Quark with GUI

living-looper-core/ - common components of C++ Living Looper plugins.

living-looper-juce/ - JUCE plugin (WIP, not functional at present)

License

Some Living Looper models incorporate RAVE, which uses the Creative Commons Attribution-NonCommercial 4.0 International license.

The plugin components including living-looper-sc are licensed under GPLv3.

As I understand it, the above affects what you can do with the source code, for example, publishing your own fork is allowed, but selling a VST based on it is not. You are free to do anything you want with the actual plugins and trained models, including making commercial art installations or music releases.