Repolode
Submit
ace-step

ace-step/ACE-Step-1.5

Open-source music generation model that runs locally, creating full songs from text prompts with editing, cover, and LoRA training features.

Open-source alternative to:SunoUdio

View on GitHubWebsite
Stars
13,790
Forks
1,816
Open issues
169
Last commit
8 days ago

What is it?

ACE-Step 1.5 is an open-source music foundation model that generates full songs from text descriptions and lyrics. It pairs a language model that plans song structure, lyrics, and metadata with a Diffusion Transformer that renders the audio, and it runs on local hardware rather than a hosted service.

Beyond text-to-music generation, the project covers cover generation from existing audio, selective repainting of sections, vocal-to-accompaniment conversion, track separation, lyric timestamp (LRC) output, and audio analysis for BPM, key, and time signature. Users can train LoRA adapters on a small set of their own songs through a Gradio interface. Duration ranges from 10 seconds to 10 minutes, with batch generation of up to 8 songs at once.

The project provides a Gradio web UI, a REST API server, a Python API, and a CLI. It supports CUDA, Apple MPS, AMD ROCm, Intel XPU, and CPU, and offers several model sizes so it can run on GPUs with different amounts of VRAM.

Key features

  • 01Generates full songs up to 10 minutes from text prompts and lyrics.
  • 02Supports cover generation, section repainting, and vocal-to-accompaniment conversion.
  • 03Trains LoRA adapters on a small set of user songs via the Gradio UI.
  • 04Outputs LRC lyric timestamps and analyzes BPM, key, and time signature.
  • 05Runs on CUDA, Apple MPS, AMD ROCm, Intel XPU, and CPU.

Use cases

  • Producing demo tracks or background music from text descriptions on a local machine.
  • Creating covers or reworking sections of existing audio clips.
  • Training a personal style LoRA from a few of your own songs.
  • Integrating music generation into an app through the REST API or Python API.

Quick start

  1. 1

    Install uv

    curl -LsSf https://astral.sh/uv/install.sh | sh
  2. 2

    Clone the repository and install dependencies

    git clone https://github.com/ACE-Step/ACE-Step-1.5.git && cd ACE-Step-1.5 && uv sync
  3. 3

    Launch the Gradio UI (models download on first run), then open http://localhost:7860

    uv run acestep

Things to know

  • Needs a CUDA GPU for good speed; the README recommends at least 6-8GB VRAM for the lighter setups and 12GB or more for LoRA training.
  • Requires Python 3.11-3.12, and ROCm on Windows needs Python 3.12 specifically.
  • The XL models need at least 12GB VRAM with offload or 20GB without it.

Topics

#music-generation#text-to-music#diffusion#lora#audio#python#gradio#text2music

FAQ

What is ACE-Step-1.5?
Open-source music generation model that runs locally, creating full songs from text prompts with editing, cover, and LoRA training features. ACE-Step 1.5 is an open-source music foundation model that generates full songs from text descriptions and lyrics. It pairs a language model that plans song structure, lyrics, and metadata with a Diffusion Transformer that renders the audio, and it runs on local hardware rather than a hosted service.
Is ACE-Step-1.5 free to use? What license does it have?
ACE-Step-1.5 is open source under the MIT license. Check the license terms before using it in commercial products.
Is ACE-Step-1.5 still maintained?
Actively maintained. The last commit to ACE-Step-1.5 was on Oct 11, 2026.
What can I use ACE-Step-1.5 for?
Producing demo tracks or background music from text descriptions on a local machine. Creating covers or reworking sections of existing audio clips. Training a personal style LoRA from a few of your own songs. Integrating music generation into an app through the REST API or Python API.
How do I install ACE-Step-1.5?
Install uv: curl -LsSf https://astral.sh/uv/install.sh | sh Clone the repository and install dependencies: git clone https://github.com/ACE-Step/ACE-Step-1.5.git && cd ACE-Step-1.5 && uv sync Launch the Gradio UI (models download on first run), then open http://localhost:7860: uv run acestep
Where can I get ACE-Step-1.5?
The source code and installation instructions are on GitHub: https://github.com/ace-step/ACE-Step-1.5

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects