Trained a LoRA on a bunch of screenshots from Minit which turned out to be really useful for reducing the palette and details in a computationally expensive but pretty impressive way. I also greatly enjoy the "MMDiT" style <code>dither</code>ing it can occasionally produce.My only prompting recommendation is to use <code>pixel art</code> in the positive prompt! 😼You can also make it hallucinate horrible 1-bit mess that looks suspiciously almost like Minit..<pre><code>This image appears to be a screenshot from the indie video game "Minit featuring an underground cave-like environment with black walls and minimal lighting.</code></pre>It wasn't explicitly trained with text in mind, but occasionally it manages some half-decent attempts with <a target="_blank" rel="ugc" href="https://huggingface.co/zer0int/CLIP-GmP-ViT-L-14">zer0int/CLIP-GmP-ViT-L-14</a>Don't forget to show me the weird stuff you make by clicking the + Add Post button down in the Gallery, and just maybe, give a thumbs up if you are up to it!

Files with SHA256: 7eb71b294a567faf051c3a65bdf6d9cd171c0a67abe011de95e46447f8e153d3