This LoRA was trained using a <a target="_blank" rel="ugc" href="https://danbooru.donmai.us/posts/11508093">single pixel art image posted by gemumeku</a>. The original image was separated into 47 individual characters, then upscaled 10× from 240×320 using nearest-neighbor interpolation to create the training images.For training, I also experimented with a modified version of <a target="_blank" rel="ugc" href="https://github.com/wraith-executioner-year-2/Anima-Standalone-Trainer/tree/feat/fix-%2341-add-came-with-lokr">beautifulrem's Anima-Standalone-Trainer with an implementation of the CAME optimizer</a>.The resulting LoRA appears to have learned the concept of pixel-perfect rendering and the proportions of the characters to some extent. When generating at higher resolutions, it tends to preserve the original character proportions while adding more detail.The pixel-perfect reproduction is not completely accurate, however. The generated pixels often extend beyond the intended grid, and diagonal lines may occasionally appear.These artifacts can be <a target="_blank" rel="ugc" href="https://github.com/KohakuBlueleaf/PixelOE#readme">removed using the </a><code>pixelOE</code><a target="_blank" rel="ugc" href="https://github.com/KohakuBlueleaf/PixelOE#readme"> node</a>: <a target="_blank" rel="ugc" href="https://files.catbox.moe/gha5rm.png">pixelOE node example</a>

@gemumeku yume nikki inspired Style

Initial.LoKr, dim32/alpha32, CAME, UNet LR 3e-05 cosine 1,000steps, 47 images.

Files with SHA256: 106df3faa88452ba1f7d8cf0dda25b9e710ca5785a5d73bd6f00ccc818d1838d