Repolode
Submit
multimodal-art-projection

multimodal-art-projection/YuE

YuE2 is an open music generation model that writes an editable melody-and-chord score from lyrics and a style prompt, then renders it as a full song with vocals.

Open-source alternative to:SunoUdio

View on GitHubWebsite
Stars
12,080
Forks
1,389
Open issues
81
Last commit
12 days ago

What is it?

YuE2 generates complete songs from lyrics and a style prompt. It first produces a melody-and-chord plan in ABC notation, which a person or agent can read and edit, and then renders that plan into audio with vocals and accompaniment at 48 kHz stereo. The model can also generate from lyrics directly without a plan.

Beyond creating songs, the project supports zero-shot covers: a source recording is transcribed to a melody score with the companion SheetSage2 model, and the song is regenerated in a new style. Edited scores can be re-rendered, so a composition can be revised step by step, for example by changing harmony, tempo, or form before generating a new recording.

The repository provides a Python pipeline with staged calls (plan, generate semantic tokens, synthesize, decode), a CLI example, and an agent skill for generation, covers, and score editing. It targets Linux with Python 3.12 and an NVIDIA GPU with BF16 support and 24 GB VRAM. Model weights are published on Hugging Face and are licensed for non-commercial use, while the code is Apache 2.0.

Key features

  • 01Generates an editable melody-and-chord score before rendering the song.
  • 02Produces vocals and accompaniment from lyrics and a style prompt.
  • 03Creates zero-shot covers from transcribed melody scores.
  • 04Re-renders user-edited ABC scores into new recordings.
  • 05Offers a staged Python API and command-line example.

Use cases

  • Writing demo songs from lyrics in a chosen genre.
  • Recreating an existing melody in a different style as a cover.
  • Experimenting with harmony or arrangement by editing the score before rendering.
  • Building music features into an app or agent workflow.

Topics

#music-generation#audio-generation#deep-learning#generative-ai#symbolic-music#voice-cloning#pytorch#ai#foundation-models#gpt

FAQ

What is YuE?
YuE2 is an open music generation model that writes an editable melody-and-chord score from lyrics and a style prompt, then renders it as a full song with vocals. YuE2 generates complete songs from lyrics and a style prompt. It first produces a melody-and-chord plan in ABC notation, which a person or agent can read and edit, and then renders that plan into audio with vocals and accompaniment at 48 kHz stereo. The model can also generate from lyrics directly without a plan.
Is YuE free to use? What license does it have?
YuE is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is YuE still maintained?
Actively maintained. The last commit to YuE was on Oct 6, 2026.
What can I use YuE for?
Writing demo songs from lyrics in a chosen genre. Recreating an existing melody in a different style as a cover. Experimenting with harmony or arrangement by editing the score before rendering. Building music features into an app or agent workflow.
Where can I get YuE?
The source code and installation instructions are on GitHub: https://github.com/multimodal-art-projection/YuE

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects