magenta/magenta-realtime
An open-weights machine learning model for real-time music generation with Python and C++ inference engines.
Velocity · 7d
+3.8
★ / day
Trend
→steady
star history
Magenta RealTime 2 is a state-of-the-art open-weights model for real-time music generation. It provides a Python library (magenta-rt) with JAX and MLX backends for inference, a C++ inference engine (magentart::core) for efficient streaming on Apple Silicon, and example applications including DAW plugins and standalone macOS apps. The model comes in two sizes (230M and 2.4B parameters) optimized for real-time audio generation.