YuE: audio generator — what it does and what you need to run it

YuE is an open-source audio generator from the M-A-P team, which the authors describe as a model for creating full-fledged songs. According to the developers, the project is intended as an open-source alternative to systems like Suno.ai. The code is published on GitHub, and the model weights are available on Hugging Face, so the tool can be studied and run on your own hardware.

What it does

The model's primary task is text generation, i.e., generating from a text description or prompt. The repository description states that YuE is positioned as a foundation model for music generation, including full songs. The project's tags include audio-generation, music-generation, voice-cloning, and style-transfers — indicating work with audio, style transfer, and voice cloning, though the authors do not provide details in the fact sheet on how exactly these capabilities are implemented.

Architecturally, the project is related to GPT and Llama approaches, suggesting the use of language models for audio processing. The authors do not specify concrete use cases or limitations on the duration or quality of the generated audio.

What you need to run it

The code is written in Python and distributed under the Apache-2.0 license. The model weights are also available under the Apache-2.0 license. The largest weight file takes up 4.6 GB, and the full set of files is 11.6 GB. This should be taken into account when planning disk space.

The authors do not specify requirements for system RAM, GPU, or CPU. The repository was created on January 23, 2025, with the last code change on June 4, 2025. The source code is available at github.com/multimodal-art-projection/YuE, the weights at huggingface.co/m-a-p/YuE-s1-7B-anneal-en-cot, and the project website is map-yue.github.io.

Who it's for

The tool will be of interest to those looking for an open-source alternative to commercial music generators and willing to work with the model directly: running it locally, studying the code, and adapting it to their own tasks. The availability of open weights and the Apache-2.0 license allows the project to be used in your own developments.

For those expecting a ready-made service with a web interface and simple setup, YuE may not be suitable: the fact sheet contains no information about a user interface, cloud access, or pre-built packages. It is also worth noting that the authors do not publish hardware requirements, so you will have to determine whether it can run on a specific machine through trial and error.

YuE is a research project with open source code and open weights, focused on generating music from text prompts. It provides access to a model capable of full songs but requires manual setup and sufficient disk space. It suits developers and researchers who value openness and the ability to customize it to their needs.