A machine learning model for exploring latent spaces of musical scores

Downloads in past


1.1.54 years ago4 years agoMinified + gzip package size for @magenta/music-vae in KB


MusicVAE Deeplearn.js API
This JavaScript implementation of MusicVAE uses Deeplearn.js for GPU-accelerated inference.
For the Python TensorFlow implementation, see the main Magenta repo.


To use in your application, install the npm package @magenta/music-vae, or use the pre-built bundle.
You can then instantiate a MusicVAE object with:
let mvae = new MusicVAE('/path/to/checkpoint')

For a complete guide on how to build an app with MusicVAE, read the Melody Mixer tutorialcl-tutorial.

Pre-trained Checkpoints

Several pre-trained MusicVAE checkpoints are hosted on GCS. While we do not plan to remove any of the current checkpoints, we will be adding more in the future, so your applications should reference the checkpoints.json file to see which checkpoints are available.
If your application has a high QPS, you must mirror these files on your own server.

Example Applications

Example Commands

yarn install to install dependencies.
yarn build to produce a commonjs version with typescript definitions for MusicVAE in the es5/ folder that can then be consumed by others over NPM.
yarn bundle to produce a bundled version in dist/.
yarn run-demo to build and run the demo.