Skip to content

erl-j/masked-generative-music-transformer

Repository files navigation

MGMT - Masked Generative Music Transformer

DOI

This project applies techniques from MaskGIT 1 / Muse 2 to symbolic music.

We currently use simple piano rolls of size 36 pitches X 32 timesteps as our target representation.

The goal of the project is to apply a pretrained model to tasks such as:

  • Scale/Rhythm constrained generation.
  • "Genre" conditioning.
  • Velocity generation..
  • Microtiming generation etc..

We are also working on applying the model to other music representations.

If you found this project useful for you research, please cite:

@software{nicolas_jonason_2023_7703864,
  author       = {Nicolas Jonason and
                  Bob L.T. Sturm},
  title        = {{erl-j/masked-generative-music-transformer: 
                   Original release}},
  month        = mar,
  year         = 2023,
  publisher    = {Zenodo},
  version      = {first},
  doi          = {10.5281/zenodo.7703864},
  url          = {https://doi.org/10.5281/zenodo.7703864}
}

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages