Magenta

by Google DeepMind

About the model:

Music generation model by Google DeepMind

Model type:

Audio

Model performance class:

Full

Link to the model:

https://huggingface.co/google/magenta-realtime

Base models:

(undefined)

End model:

Magenta RealTime

End model license:

Apache-2.0 and CC-BY-NC-4.0 and additional terms

About the organisation:

Major technology company, operator of Google Search.

Link to the organisation:

https://deepmind.google/

Model release date:

June 2025

Availability

Base Model Data

No details provided for training data.

End User Model Data

No details found for training data.

Base Model Weights

Model available through HuggingFace.

https://huggingface.co/google/magenta-realtime

End User Model Weights

Model available through HuggingFace.

https://huggingface.co/google/magenta-realtime

Training Code

No pre-training or instructing-tuning code made available. Some developer tools available.

https://github.com/magenta/magenta-realtime

Documentation

Code Documentation

No pretraining or finetuning code found. Scarce code documentation except for deployment of the open weights.

Hardware Architecture

Processors used are described, for the rest scarce info is provided.

https://huggingface.co/google/magenta-realtime

Preprint

Model published on arXiv.

https://arxiv.org/abs/2508.04651

Paper

Published as poster in NeurIPS.

https://neurips.cc/virtual/2025/loc/san-diego/129287

Modelcard

Model card provides the requisite information.

https://huggingface.co/google/magenta-realtime

Datasheet

No datasheet found.

Access

Licenses

Apache-2.0 and CC-BY-4.0, with additional terms of use

https://huggingface.co/google/magenta-realtime

Is this information not up to date?

Contribute here ->

Supported by the Centre for Language Studies and the Dutch Research Council. Website design & development © 2024 by BSTN. This version of the index generated 09 April 2026, website content last updated 11 March 2026.