What can you build with ACE-Step-1.5?

Generate a full song with lyrics from a one-line description. Make a cover of an existing track in a different style. Repaint a weak section of a generated song without redoing the whole thing. Train a LoRA on your own songs so new tracks match your style. Add generated music to an app through the REST or Python API.

ace-step /

ACE-Step-1.5

Open-source music generation model that runs on your own computer

13.9k stars
Python
Updated 11 days ago
Some experience
Image, video & audio

Open on GitHub (opens in a new tab) Website (opens in a new tab)

ACE-Step 1.5 generates music from text. You describe a song, or give lyrics and style tags, and it produces audio from 10 seconds up to 10 minutes long. It runs locally, so you don't need a paid music service. The project also offers a free hosted version at acemusic.ai for people without a GPU.

It works in two stages. A language model plans the song first, expanding a short prompt into a blueprint with metadata, lyrics and a caption. A diffusion transformer (DiT) then turns that plan into audio. The README says quality lands between Suno v4.5 and Suno v5 on its own metrics. Those are the authors' claims, so listen to the output yourself.

Beyond plain generation, it edits audio. It can make covers, repaint sections of a track, split stems, and add accompaniment to a vocal. You can also train a LoRA on a few of your own songs to capture your style. Model sizes run from 2B to 4B (XL), so you can pick one that fits your GPU memory.

What it does

Text to full song
Turns a short description into a song with lyrics. Lyrics in 50+ languages are supported. You can set duration, BPM, key and time signature.
Runs on consumer hardware
Supports CUDA, Apple Silicon, AMD ROCm, Intel XPU and CPU. The README says the smallest setup needs under 4GB of VRAM, and a table matches your VRAM to a model.
Editing tools
Includes cover generation, repainting selected sections, track separation, multi-track layering and Vocal2BGM. Reference audio can guide the style.
LoRA training
Train a personal style from a few songs inside the Gradio UI. The README gives 8 songs in about 1 hour on a 3090 as an example.
Several ways to use it
Comes with a Gradio web UI, a REST API, a Python API, a CLI, and a separate VST3 plugin for DAWs.

What you could build

Generate a full song with lyrics from a one-line description.
Make a cover of an existing track in a different style.
Repaint a weak section of a generated song without redoing the whole thing.
Train a LoRA on your own songs so new tracks match your style.
Add generated music to an app through the REST or Python API.

Get started

Install uv, clone the repo and run uv sync. Then start the Gradio UI with uv run acestep, which downloads the models on first run, and open http://localhost:7860. Windows and macOS portable packages are also available.

curl -LsSf https://astral.sh/uv/install.sh | sh

git clone https://github.com/ACE-Step/ACE-Step-1.5.git
cd ACE-Step-1.5
uv sync

uv run acestep

Commands are copied from the project's README. Read it on GitHub for the full setup.

Who it's for

Suits musicians, hobbyists and developers who want music generation on their own machine. You should be comfortable with a terminal and know roughly how much GPU memory you have.

Useful for

Creators and marketersGenerate full songs with lyrics locally, make covers and repaint weak sections.
ML engineers and data scientistsOpen music generation model running across Mac, AMD, Intel and CUDA to study or fine-tune.
Game developersGenerate original background music and tracks locally for games.