ACE Step

ACE Step

Generates music from text descriptions and lyrics with the ACE-Step foundation model, in a browser demo. It suits musicians, producers and content creators who want fast AI music drafts in many styles and languages.

Made by ACE-Step 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 is a demo of ACE-Step, an open-source foundation model for music generation. The authors combine diffusion-based generation with a compression autoencoder and a light linear transformer, aiming for speed, musical coherence and control. They say it makes up to 4 minutes of music in 20 seconds on an A100 GPU. The demo runs on zero-a10g hardware at Hugging Face.

Who it suits: It suits people who make music or need background tracks and want quick drafts from a text description, and researchers who want a base model for music tasks.

Industry | Who uses it | How they use it | Example
Music production | Producers and songwriters | Draft tracks from a style description and lyrics | Describe a genre and mood, add lyrics and listen to the draft
Content creation | Video and podcast makers | Make background music for a project | Ask for an instrumental piece that fits a scene
AI research | Researchers | Build sub-tasks on a music foundation model | Use the model as a base for further training
Education | Students | Hear how different styles and instruments sound | Compare the same prompt in two genres

You can skip it if you need finished, mixed releases, since the output is a generated draft.

Where to use it: You can use it in these places:

• A browser: open the Space's page on Hugging Face and nothing needs installing.
• Windows, macOS or Linux: run your own copy with Python.

Use case | Who | What they get back
Text to music | Musicians | A generated song or instrumental from a description
Variations | Producers | Alternative versions of a generated piece
Voice and lyric work | Songwriters | Results from voice cloning, lyric editing and remixing, which the authors list as supported

How to set it up: You need these first for a local copy:

• Python and pip
• The system packages listed in packages.txt, which Hugging Face installs with apt-get
• A GPU able to run a 3.5B model

The online page needs none of this. For your own copy, follow the steps below.

1. Install the Hugging Face tools.
2. Download the Space's files into a folder.
3. Move into that folder.
4. Install Gradio at the version the Space uses, then the other packages.
5. Start the app.

pip install -U huggingface_hub
hf download ACE-Step/ACE-Step --repo-type space --local-dir ACE-Step
cd ACE-Step
pip install gradio==5.27.0
pip install -r requirements.txt
python app.py

How to use it: Open the Space in a browser, or start your own copy with the command below. Then describe the music you want, add lyrics if you like and run the generator.

python app.py

Example: You describe a style in a few tags, such as a genre and a mood, and paste in some lyrics. The Space returns a generated track that follows the style and sings the lyrics.

For newcomers and experts: If you are new to this, start with a short style description and no lyrics. Listen to the instrumental, then add lyrics and run it again.

If you know the field, the design joins a diffusion model with a deep compression autoencoder and a linear transformer, and it aligns semantic features during training for quicker convergence. The README says it supports 19 languages and names ten that work best, among them English, Chinese and Spanish. Less common languages may do worse because of uneven data.

You write a style description
You add optional lyrics
The model generates the music
You listen and make variations

Good to know: The README lists training code, a LoRA release and ControlNet code as still to come. The Space also names a Chinese rap LoRA.

• License: apache-2
• Cost to run: none named
• Data it reaches: the prompts and lyrics you enter, and model files downloaded from Hugging Face

On Hugging Face, 13 November 2026

• Likes: 692
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/ACE-Step/ACE-Step
• Files: https://huggingface.co/spaces/ACE-Step/ACE-Step/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.