---
title: YuE2 Studio
emoji: 🎵
colorFrom: indigo
colorTo: purple
sdk: gradio
sdk_version: 5.49.1
app_file: app.py
python_version: "3.10.13"
short_description: Turn lyrics into songs and edit their musical scores
startup_duration_timeout: 1h
models:
  - m-a-p/YuE2-3B
  - m-a-p/YuE2-Vae
  - Comfy-Org/YuE2
---

# YuE2 Studio

Lyrics-to-song generation, editable ABC conditioning, FLAC playback/downloads,
reproducible seeds, settings export, and a Gradio API/MCP endpoint.

## Model provenance

The linked **Comfy-Org/YuE2** repository repackages YuE2 for ComfyUI.
This Gradio application uses the original **m-a-p/YuE2-3B** Python inference
package and **m-a-p/YuE2-Vae**, not ComfyUI's single-file checkpoint.
The included official ComfyUI text-to-music workflow supports the Comfy-Org files.

Sources:
- https://huggingface.co/Comfy-Org/YuE2
- https://huggingface.co/m-a-p/YuE2-3B
- https://github.com/Comfy-Org/ComfyUI/blob/master/blueprints/Text%20to%20Music%20%28YuE2%29.json
- ZeroGPU loading reference: https://huggingface.co/spaces/mrfakename/yue2-3b

Model weights are CC BY-NC 4.0. Respect attribution and noncommercial restrictions.

## Deploy

Requires an authenticated Hugging Face account eligible for ZeroGPU.

```sh
hf repos create adexinc/yue2-studio --type space --space-sdk gradio --flavor zero-a10g --public
hf upload adexinc/yue2-studio . --repo-type space --exclude '**/__pycache__/**'
```

Run these commands from this folder. No paid dedicated GPU is requested.
The upstream inference wheel pins torch 2.10 and its matching dependencies.
Platform-managed Gradio and spaces are intentionally omitted from requirements.
The GPU timeout is provisional until live inference measurements are available.

## Validation status

Deployed on ZeroGPU and tested through the live Gradio API. The smoke test
generated 46.3 seconds of 48 kHz stereo FLAC audio plus an editable ABC score
and settings bundle in 23.2 seconds. The returned FLAC header and runtime logs
were checked after generation.