ACE-Step
Open-source AI model that generates full songs from text prompts and runs on your own computer.ace-step/ACE-Step-1.5
- Stars
- 14.3k
- Forks
- 1.9k
- Licence
- MIT
- Language
- Python
- Last update
- 25 days ago
- Type
- AI model / agent
What it is
ACE-Step 1.5 is an open-source music generation model. You describe a song in text, and it can produce audio, including lyrics and structure, that runs locally on your own hardware instead of a cloud service. The README also lists editing features such as covers, repainting sections, and separating audio into stems. People use it to experiment with music creation without relying on a paid commercial service.
- Generate a full song from a short text description with lyrics and style tags
- Create a cover or repaint a selected section of an existing audio clip
- Generate accompaniment for a vocal track (Vocal2BGM)
- Train a LoRA on a small set of your own songs to capture a personal style
- Extract BPM, key, and time signature from an audio file
Great for
- Musicians and producers who want a free, local tool for song sketches and ideas
- Creators who need background music or accompaniment generated from vocal tracks
- Developers who want to build music features through its Gradio UI or REST API
- Users with a CUDA, Apple Mac (MPS), AMD (ROCm), Intel, or CPU setup who want to run generation on their own machine
Not ideal for
- Anyone without a reasonably capable GPU or patience for setup, since the README recommends a CUDA GPU and the full guide covers several hardware paths
- Users who want a polished consumer app with no installation; the README points to the hosted acemusic.ai site for that
- Projects needing guaranteed commercial licensing of outputs, since the README does not address output rights beyond the MIT license of the code
How to get it
- 1
Install with a package manager
You'll need: Python 3.11-3.12 · uv (Python package manager) · CUDA GPU recommended; MPS, ROCm, Intel XPU, or CPU also supported per README
Install uv (macOS / Linux)
curl -LsSf https://astral.sh/uv/install.sh | shClone the repository
git clone https://github.com/ACE-Step/ACE-Step-1.5.gitInstall dependencies
cd ACE-Step-1.5 && uv syncLaunch Gradio UI
uv run acestep - 2
Run it
- Install uv and clone the repository, then run uv sync inside the project folder
- Start the Gradio UI with uv run acestep; models download automatically on first run
- Open http://localhost:7860 in your browser
- Enter a text description or lyrics and generate a track
- 3
Get the most out of it
- Use the model-selection table in the README to pick a DiT and LM size that matches your GPU VRAM
- Try the Simple Mode and Query Rewriting features to let the LM expand short prompts into fuller song blueprints
- Use LoRA training in Gradio with a small set of your own songs to personalize output style
- Use uv run acestep-api to run the REST API server on port 8001 for integration into other tools
- Check the Installation Guide in the docs folder for environment variables and command-line options
Downloads
Strengths
- Runs locally on several hardware types, with documented options for lower VRAM setups
- Offers a wide range of editing and control features beyond basic text-to-music
- Released under the MIT license with a large community (over 14,000 GitHub stars)
Trade-offs
- Setup varies by hardware and can require Python 3.11-3.12 plus specific GPU drivers; the full install guide is needed for AMD, Intel, or CPU
- The README's performance claims (speed, quality versus commercial models) are self-reported and not independently verified here
- Larger XL models need at least 12GB VRAM with offload, or 20GB without it, which excludes many laptops
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.