Magenta RealTime
Google Magenta's first real-time music generation open model
Google Magenta's first real-time music generation open-source model.
Generate music and songs with open AI models.
Key features
- Open-weights real-time streaming music generation model
- Python library magenta-rt with JAX and MLX backends
- C++ inference engine for Apple Silicon; AUv3 DAW plugin
- Two sizes: 230M small and 2.4B base
- Example apps: standalone, jam, collider
Good for
- Real-time generative music on Mac
- Building DAW plugins and music apps
Stack & tags
PythonJAXC++music-generationreal-timestreamingopen-weightsmlxjaxapple-silicondaw
Quick start
mrt mlx generate --prompt "disco funk" --duration 4.0 --model=mrt2_baseFrom the project README on GitHub.
Similar open-source projects
ACE-Step-1.5
★ 13,569The most powerful local music generation model that outperforms almost all commercial alternatives, supporting Mac, AMD, Intel, and CUDA devices.
ACE-Step
★ 4,868ACE Studio + StepFun music foundation model — 4 minutes of music from 20 seconds