I’ve been using cloud-based image generation for a while—paying every month, and on top of that, all my prompts and images get uploaded to someone else’s servers. That always felt a bit off. Recently I found an open-source desktop app that runs Illustrious XL v2.0 locally on your own machine, supports both Mac and PC, so I gave it a shot.
This model is an SDXL variant, specifically tuned for anime and illustrations. It uses danbooru-style tag prompts—comma-separated tag descriptions, not long sentences. It feels a bit weird at first if you’re coming from other models, but after a few minutes you actually find it faster, like picking attributes from a menu. Quality tags usually go at the end, and tags earlier in the prompt have higher weight.
It uses the GGUF quantized version, so both download size and VRAM usage are smaller than the original. On Mac it runs via Metal, on Windows via CUDA or Vulkan. If you don’t have a GPU, you can still run it on CPU, just slower. There are several quantization levels to choose from—if you’re tight on memory, go for a more aggressive one. For anime with flat shading and line art, even a small quant can handle it. The whole thing has no accounts, no telemetry—your images and prompts stay in a local folder. The code is AGPL open-source, so you can check it yourself.
