Magenta
Google research project using ML (TensorFlow) to generate music, art, and drawings.
| Language | Python |
|---|---|
| Category | ML & generative |
| License | Apache-2.0 |
| Platforms | macOS Windows Linux |
| Install | pip install magenta |
| First released | 2016 |
| Maintained | Not actively maintained (last activity 2026) |
| Links | Website · Source |
Strengths
- Broad set of pretrained music/art models (MusicVAE, MelodyRNN)
- Backed by Google Brain research with strong papers
Limitations
- Main repo archived Jan 2026, now read-only
- Aging TensorFlow 1.x-era stack
Best for
Learning classic deep-learning approaches to symbolic music generation and reusing pretrained models.
Minimal example
from magenta.models.music_vae import TrainedModel
# load a pretrained checkpoint and sample new sequences