acestep.cpp ServeurpersoCom/acestep.cpp

C++/GGML port of ACE-Step 1.5 with a local web UI and API; runs quantised song models on CPU, CUDA, Metal or Vulkan without Python.

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

acestep.cpp at a glance

Checked 15 Oct 2026

acestep.cpp reimplements ACE-Step 1.5 in C++ on GGML and serves a browser UI and HTTP API on your machine. It loads GGUF quantisations (a default set is about 7.7 GB at Q8) and runs on CPU, NVIDIA CUDA, Apple Metal or AMD/Intel via Vulkan. It supports text-to-music, cover, repaint and LoRA adapters. Code and GGUF weights are MIT. You build it from source; Windows binaries are on a temporary download page until CI is set up. No RAM or VRAM figures are given. Several desktop apps (ACE-Step Studio, HOT-Step) are built on it.

Cite this pageRepoQuarry. “acestep.cpp.” Checked 15 Oct 2026. https://repoquarry.dev/projects/acestep-cpp

How much of Suno does it replace?

3 of 5 jobs. Songs, covers and repaints from a local web UI; no built-in stem tool documented.

Full songs with vocals from a prompt or lyrics Partial replacement

Instrumental tracks (Instrumental mode) Not covered

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) Not covered

For instrumental tracks (Instrumental mode), splitting a song into vocal and instrument stems (Get Stems), see other Suno alternatives. You may also need to build an interface, an API server and hosting yourself.

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, local server with a browser UI Describe a song, get stereo 48kHz audio. Runs on CPU, CUDA, Metal, Vulkan. github.com · acestep.cpp#readme ↗ 1d ago

Code licence MIT github.com · LICENSE ↗ 1d ago

Model weights licence MIT (GGUF conversions of ACE-Step 1.5) huggingface.co · ACE-Step-1.5-GGUF ↗ 1d ago

Commercial use Yes, MIT code and weights huggingface.co · ACE-Step-1.5-GGUF ↗ 1d ago

Paid services needed? None required github.com · acestep.cpp#readme ↗ 1d ago

Hardware Runs on CPU or GPU; default Q8 model set is about 7.7 GB on disk; no RAM or VRAM figures Grab one GGUF of each type from Hugging Face and drop them in the models/ folder github.com · acestep.cpp#download-models ↗ 1d ago

Operating systems Windows, Linux and macOS (Metal) macOS auto-enables Metal and Accelerate BLAS with any of the above. github.com · acestep.cpp#build ↗ 1d ago

Setup effort Hard; build from source with CMake scripts (CUDA/Vulkan SDK optional); unofficial Windows binaries Pre-built binaries (until CI is set up) github.com · acestep.cpp#build ↗ 1d ago