Di♪♪Rhythm

Di♪♪Rhythm

Generates songs with the DiffRhythm model from the ASLP lab, in a Gradio Space on GPU hardware. For musicians, researchers and curious users who want to hear AI-made songs quickly.

Made by ASLP-lab on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space lets you try DiffRhythm, a song generation model from the ASLP lab, behind a Gradio page. It names several model files, including DiffRhythm 1.2, a full version, a base version and a VAE, plus two MuQ music models. The README is one line, so the controls are not described there.

Who it suits: It suits musicians, producers and researchers who want to hear what a current AI song generator can do. It also suits developers who want to read how a larger music model is wired into a Gradio app.

Industry | Who uses it | How they use it | Example
Music | Songwriters | Hear a rough AI take on a song idea | Run the generator and listen to the result
Media | Video creators | Look for placeholder music for a draft | Generate a track and test it against a cut
Research | Audio researchers | Try a published song model | Compare output across model versions
Software | Developers | Study a music model served through Gradio | Open app.py and the diffrhythm folder

Anyone who needs licensed, release-ready music can skip it, since this is a research demo and the page gives no usage rules.

Where to use it: The demo runs in two places.

• A web browser, on the page of the Space at Hugging Face
• Windows, macOS or Linux, as a local copy run with Python

Use case | Who | What they get back
Song generation | Musicians | A generated song to listen to
Model comparison | Researchers | A way to hear the base, 1_2 and full versions
Code study | Developers | A repository with the app, the diffrhythm package and pretrained files

How to set it up: The online page needs nothing installed. A local copy needs the following.

• Python and pip
• Windows, macOS or Linux
• The packages listed in requirements.txt
• The system packages listed in packages.txt, which the Space installs with apt-get
• A way to fetch the DiffRhythm and MuQ models the Space names

The Space itself runs on ZeroGPU hardware with an A10G. The source does not say what a local copy needs.

1. Install Python.
2. Install the Hugging Face hub package and download a copy of the Space.
3. Move into the new folder and install Gradio 5.20.0 and the packages in requirements.txt.
4. Start the app with python app.py.

pip install -U huggingface_hub
hf download ASLP-lab/DiffRhythm --repo-type space --local-dir DiffRhythm
cd DiffRhythm
pip install gradio==5.20.0
pip install -r requirements.txt
python app.py

How to use it: Open the page, fill in the inputs it asks for, run it, and play the song it returns. On a local copy, this command starts the page.

python app.py

Tool | What it does
DiffRhythm | The song generation model, in a 1_2 version, a full version and a base version
DiffRhythm-vae | The audio autoencoder the Space names
MuQ-large-msd-iter | A MuQ music model the Space names
MuQ-MuLan-large | A second MuQ model the Space names
app.py | Holds the Gradio page

Example: You open the page, supply the inputs it asks for, and run the generator. After a wait, it returns a generated song that you can play in the browser.

For newcomers and experts: If you are new to this, start with the default inputs on the page and listen to one result. Then change one input at a time and compare the songs.

If you know the field, the Space bundles several DiffRhythm checkpoints and two OpenMuQ models, with the diffrhythm package, a LangSegment folder, a pretrained folder and an src folder in the repository. It runs on ZeroGPU with an A10G, so waits come from the shared GPU. The page documents no parameters, so read app.py for the controls.

Fill in the inputs
DiffRhythm generates the song
The VAE decodes the audio
The song plays in the page

Good to know: The Space page gives one line of description. Settings and limits are best read from app.py.

• License: Apache 2
• Cost to run: none named
• Data it reaches: the inputs you give it, and the models the Space names
• Hardware: ZeroGPU with an A10G at Hugging Face

On Hugging Face, 13 November 2026

• Likes: 690
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/ASLP-lab/DiffRhythm
• Files: https://huggingface.co/spaces/ASLP-lab/DiffRhythm/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.