# magenta/magenta-realtime

Magenta RealTime 2: An Open-Weights Live Music Model

Repository: https://github.com/magenta/magenta-realtime
Canonical: https://ross.abutalabs.com/products/magenta-realtime
Homepage: https://magenta.withgoogle.com/magenta-realtime-2
Language: Python
License: Apache-2.0
License Family: permissive
Topics: jax, machine-learning, mlx, music, real-time
Last push: 2026-08-24T21:54:44+00:00

## Health v2 (maintenance only)
Score: 73/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 63, longevity 31
- inputs: {"age_days": 447, "days_push": 9, "days_rel": 34, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1739, forks 192 (observed 2026-08-28T04:05:30.070061+00:00)

## What it is
Magenta RealTime 2 (MRT2) is an open-weights model for real-time, interactive music generation controllable via text, MIDI, and audio. It ships with a Python inference library (JAX/MLX), a C++ streaming inference engine for Apple Silicon, and example apps including a DAW plugin.

## Use cases
- generate music in real time on a MacBook
- build an AI musical instrument controlled by MIDI
- run live music generation inside a DAW
- stream audio faster than playback on Apple Silicon
- run offline music generation on an NVIDIA GPU
- clone sounds and blend styles interactively
- embed a live music model into custom music software

## When to choose
- you need low-latency, interactive music generation rather than offline track generation
- you are on Apple Silicon and want on-device inference
- you want open weights you can fine-tune or embed in your own apps
- you want a DAW plugin or standalone app for live performance

## When to avoid
- you need real-time streaming on non-Apple hardware (only offline inference is supported there)
- you want text-to-song generation of finished tracks rather than live performance
- you need supervised fine-tuning today (planned but not yet supported)

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, audio-processing, llm-inference, sdk
- domain: machine-learning, developer-tools
- platform: python, cpp
- tags: music-generation, real-time-audio, open-weights, jax, mlx, audio-plugin, daw, generative-music, apple-silicon, streaming-inference, audio, real-time, macos, gpu

## Member repositories
- magenta/magenta-realtime (main) score 73

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:30.070061+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:30:13.322179+00:00, confidence not recorded.
  - readme: https://github.com/magenta/magenta-realtime (fetched 2026-08-28T04:05:30.070061+00:00, sha d2d61589a473)
  - homepage: https://magenta.withgoogle.com/magenta-realtime-2 (fetched 2026-08-29T11:07:36.492634+00:00, sha bb67020fb1b1)
- Data as of 2026-08-30T08:39:29.467469+00:00.