ACE-Step 1.5 ace-step/ACE-Step-1.5

Open-weight music model that writes and renders full songs with vocals locally, with covers, repainting, stem extraction and LoRA training.

Creator kit + Compare GitHub ↗ Install guide (official) ↗

ACE-Step 1.5 at a glance

Checked 15 Oct 2026

ACE-Step 1.5 (co-led by ACE Studio and StepFun) pairs a language-model planner with a diffusion decoder to make songs from 10 seconds to 10 minutes, with lyrics in 50+ languages. It runs as a Gradio UI, REST API or CLI, and the 2B model fits in under 4 GB of VRAM; the 4B XL models need 12 GB or more. Code and weights are MIT. Covers, repaint and stem extraction exist, but extraction needs the base models. The YuE2 team's benchmark scores it below Suno v5, and a source install needs Python 3.11-3.12 unless you use the portable packages.

Cite this pageRepoQuarry. “ACE-Step 1.5.” Checked 15 Oct 2026. https://repoquarry.dev/projects/ace-step-1-5

How much of Suno does it replace?

5 of 5 jobs. Covers all five jobs locally; stem extraction only with the base models. Output quality was not tested here.

Full songs with vocals from a prompt or lyrics Partial replacement

Instrumental tracks (Instrumental mode) Partial replacement

Covers and restyles of an uploaded song (Cover) Partial replacement

Extending or regenerating part of a song (Extend Partial replacement

Splitting a song into vocal and instrument stems (Get Stems) Partial replacement

Facts, with sources

Every answer links to where we found it. We read the documentation; we did not install it.

Report a correction

QuestionAnswerSourceChecked

Runs on my machine? Yes, Gradio UI, REST API or CLI on your own machine The model runs locally with less than 4GB of VRAM, and supports lightweight personalization github.com · ACE-Step-1.5#-abstract ↗ 1d ago

Code licence MIT This project is licensed under MIT github.com · ACE-Step-1.5#readme ↗ 1d ago

Model weights licence MIT (ACE-Step 1.5 and XL weights on Hugging Face) huggingface.co · Ace-Step1.5 ↗ 1d ago

Commercial use Yes, MIT code and weights; the authors ask you to check outputs for copyright problems we encourage users to verify the originality of generated works, clearly disclose AI involvement github.com · ACE-Step-1.5#-license--disclaimer ↗ 1d ago

Paid services needed? None required; a free hosted version (acemusic.ai) is optional Don't want to install locally? Try acemusic.ai — 100% free, no GPU required! github.com · ACE-Step-1.5#-quick-start ↗ 1d ago

Hardware Under 4 GB VRAM for the 2B model without the planner; XL needs 12 GB+ with offload or 20 GB+ without; CUDA, ROCm, Intel XPU, Apple MPS or CPU CUDA GPU recommended (also supports MPS / ROCm / Intel XPU / CPU) github.com · ACE-Step-1.5#-which-model-should-i-choose ↗ 1d ago

Operating systems Windows (CUDA or ROCm), Linux and macOS (Apple Silicon) launch scripts Ready-to-use launch scripts for all platforms with auto environment detection, update checking, and dependency installation. github.com · ACE-Step-1.5#-launch-scripts ↗ 1d ago

Setup effort Easy; portable packages for Windows and macOS, or uv sync then uv run acestep For Windows users, we provide a portable package with pre-installed dependencies github.com · INSTALL.md ↗ 1d ago