MuseGAN
salu133445/museganReviewed by AI
Checked by RepoPicks Radar on
What is MuseGAN?
MuseGAN is the code behind a 2018 research paper on music generation. It trains a model on a piano-roll dataset and then writes short multi-track phrases (bass, drums, guitar, piano, strings), which come out as numpy arrays, images and piano-roll files you convert to MIDI yourself; there is no audio and no interface. It is for researchers and developers on Linux with Python and a graphics card, and the pinned packages are the 2018 versions of TensorFlow 1 and numpy, so the environment has to be rebuilt by hand.
MuseGAN is free and open source (MIT) and can replace Suno and Udio.
How easy is it to install MuseGAN?
E · developers only · Linux
Linux is the only documented path and it is developer work: every step is a shell script (download_data.sh, run_inference.sh), and requirements.txt pins tensorflow-gpu==1.10.1, numpy==1.14.5 and setuptools==39.1.0 from 2018, which no longer install on current Python without rebuilding an old environment.
Coming soon: install in 1 click with the RepoPicks app.
MuseGAN on GitHub
- Stars
- 2,045on Nov 10, 2026
- Downloads
- —all releases
- Growth
- —no gain in the last reading
Stars over time
Facts
- Runs on
- Linux
- You need to install
- Python, Graphics card
- To build from source
- Python, Graphics card
- Disk space
- 100 MB–300 MB · estimate
- Accounts or keys
- None
- Cost
- Free
- License
- MIT (Permissive)
- Last change on GitHub
- Jun 7, 2024
MuseGAN risk signals (6)
- AttentionDependencies with known vulnerabilities
Pinned dependencies with known vulnerabilities (OSV): setuptools: GHSA-5rjg-fvgr-3xxf, GHSA-cx63-2mw6-8hw5, GHSA-h35f-9h28-mq5c; grpcio: GHSA-496j-2rq6-j6cc, PYSEC-2026-1424; markdown: GHSA-5wmx-573v-2qwq, PYSEC-2026-89; numpy: GHSA-5545-2q6w-2gh6, GHSA-6p56-wp2h-9hxr, GHSA-9fq2-x9r6-wfmf; pillow: GHSA-254f-3936-288h, GHSA-27h8-gf7j-v5hc, GHSA-332p-7hrc-92wf; protobuf: GHSA-7gcm-g887-7qv7, GHSA-8gq9-2x98-w8hf, GHSA-8qvm-5x2c-j2w7; pyyaml: GHSA-8q59-q68h-6hv4, GHSA-rprw-h62v-c2w7, PYSEC-2018-49; scipy: PYSEC-2023-102, PYSEC-2023-114; tensorflow-gpu: GHSA-23hm-7w47-xw72, GHSA-2475-53vw-vp25, GHSA-247x-2f9f-5wp7; werkzeug: GHSA-29vq-49wr-vm6x, GHSA-2g68-c3qc-8985, GHSA-87hc-h4r5-73f7
automatic check
- AttentionLow OpenSSF Scorecard
OpenSSF Scorecard 2/10
automatic check
- Attentionobsolete framework pinned
requirements.txt pins tensorflow-gpu==1.10.1, tensorboard==1.10.0, numpy==1.14.5 and scipy==1.1.0, the 2018 releases; TensorFlow 1.x needs an old Python and CUDA version and is no longer maintained.
AI review
- Infodownloads data from outside github
README.md's `./scripts/download_data.sh` and `./scripts/download_models.sh` pull the training data and pretrained models from a personal OneDrive account (ucsdcloud-my.sharepoint.com), not from the repository.
AI review
- Infono audio output
The Outputs section lists .npy, .png and .npz piano-roll files and shows Python code using Pypianoroll to write a .mid file: there is no audio render and no player.
AI review
- Infono recent activity
Last push 2024-06-07 (885 days before this check), no releases, 21 open issues, and the README points to another repository for a PyTorch version.
AI review
Evidence is shown in its original language.
What Radar could not check about MuseGAN
- whether the OneDrive links for the dataset and the pretrained models still work
- how the generated phrases actually sound once turned into MIDI and played with instruments
- whether the old TensorFlow 1 code can be made to run on a current graphics card
- how long training takes and how much disk the dataset needs