Hi, I’m Hal The app · coming soon ♥ Donate
Menu
Free port · by Hi, I’m Hal

YuE2 music generation, running on AMD.

A working port of the YuE2 AI music generator for AMD Strix Halo and Radeon 8060S. No NVIDIA card, no CUDA.

Download — coming soonSetup stepsSkip the setup

The port is in final testing on Hal’s own AMD machine — the download lands here shortly. This port is free. Support the project. · YuE2 is licensed non-commercial (CC BY-NC). This port is free.

Made with this port

Hear it: “Bet on Me”

Written by Gary Gardner and generated start to finish with YuE2 on AMD — a BOSGAME M5 mini PC with a Ryzen AI Max+ 395 (Strix Halo). No NVIDIA card, no cloud. This is what the port sounds like.

“Bet on Me” · 3:08 · Lyrics by Gary Gardner · Music generated with the YuE2 AMD port

Lyrics

[Verse] chance has no chance as the odds slip away chance has no chance unless you play [Chorus] bet on me ’cause i bet on you double down double down bet on me ’cause i bet on you double down double down

Requirements

  • Tested hardwareRyzen AI Max+ 395 (Strix Halo), Radeon 8060S
  • Tested onBOSGAME M5 mini PC
  • Operating systemConfirm OS + version
  • ROCm versionConfirm
  • PythonConfirm
  • Memory neededConfirm from tests
  • Disk for weightsConfirm

Before you start

Setup takes a working ROCm install, a Python environment and a large model download. If you are comfortable in a terminal, follow the steps below. If not, use One Click Local AI or book a setup session.

One Click Local AIPaid setup help
Step by step

Install in five steps.

Commands are shown as an outline for now. The full step-by-step will be published here with the release.

  1. Install AMD drivers and ROCm

    Install the AMD graphics driver and the ROCm release supported by the port.

    # PLACEHOLDER: exact ROCm install command for the tested OS

  2. Clone the AMD port

    git clone https://github.com/PLACEHOLDER-USER/PLACEHOLDER-YuE2-AMD.git
cd PLACEHOLDER-YuE2-AMD

  3. Create an environment and install requirements

    python -m venv .venv
# activate it, then:
pip install -r requirements.txt   # PLACEHOLDER: confirm file name / ROCm wheels

  4. Download the YuE2 weights

    Get the model weights from the official YuE2 source and accept its non-commercial licence. The port does not redistribute weights.

    # PLACEHOLDER: official weights download command / link

  5. Generate your first song

    # PLACEHOLDER: example generation command from the README

    Stuck? Bug reports will live on GitHub once the port is released — until then, see setup help.

FAQ

Questions from AMD folks.

Is the YuE2 AMD port free?

Yes — the port is free, and YuE2 is licensed CC BY-NC (non-commercial). The port is still in final testing and isn’t posted for download yet; when it’s released it will be a free extra from this page. If it helps you, support options (Ko-fi and GitHub Sponsors) will open here once those accounts are live.

Can I use YuE2 or its music commercially?

YuE2 is released under a non-commercial licence (CC BY-NC). Read the original YuE2 licence before using it or its output for anything commercial. Hi, I'm Hal does not sell YuE2 or YuE2 output.

Which AMD hardware has been tested?

AMD Ryzen AI Max+ 395 (Strix Halo) with Radeon 8060S graphics, on a BOSGAME M5 mini PC. Other AMD GPUs may work but have not been tested yet.

Does it need an NVIDIA GPU or CUDA?

No. That is the point of the port: it runs on AMD hardware using AMD's ROCm software stack instead of CUDA.

How long does a song take to generate?

Timing results from Hal's own tests will be published here. Until then, no number is quoted rather than guessing.

Something broke. Where do I get help?

Once the port is released, bug reports will live on its GitHub repo. Until then, reach Gary through the Support page.

What is One Click Local AI, and is YuE2 part of it?

One Click Local AI is Hal's separate paid Windows app for installing local AI in one click. It is not a paid version of YuE2. YuE2 stays free and is governed by its own licence.

Support the port

Kept free by sponsors.

Porting and testing on AMD takes real hardware and many evenings. If this saved you time, chip in.

GitHub Sponsors Ko-fi Buy Me a Coffee
Placeholder links · real URLs coming

Credits

YuE2 is the work of its original authors and is licensed CC BY-NC. The AMD port is by Hi, I’m Hal. Add upstream repo + author links