Player not loading? Watch on YouTube
This review and setup tutorial covers ACE-Step 1.5 XL, an open source music generator the presenter runs locally on Windows. Music examples include Italian opera, Spanish Latin trap, Japanese pop and instrumental tango. The presenter reports cleaner audio and better vocals than the earlier version. Claims that XL outperforms Udio and Suno v5 come from the project's benchmarks, rather than an independent comparison in this tutorial.
The walkthrough installs uv and Git, clones the repository, installs dependencies and downloads XL Turbo through the Hugging Face CLI. The presenter describes the download as roughly 20 GB and notes support for AMD GPUs and Apple silicon, though the demonstrated setup uses Windows.
For the official XL models, the stated minimum is 12 GB VRAM with CPU offload and INT8 quantization; 20 GB is recommended for fitting the model entirely on the GPU. The optional language model for thinking mode needs additional memory, with the presenter suggesting 24 GB to be safe. Turbo uses 4 to 8 steps, while SFT uses 30 to 50 and is described as slower but better in quality.
The interface walkthrough covers initialization, style prompts, lyrics and batch generation. The presenter says tempo, key and time-signature controls do not always work well. Reference audio, section editing and remixing receive a brief overview. The tutorial presents offline generation after installation and model downloads.
