Makes fantasy images with a dreamy touch similar in style to the artist <a rel="ugc" href="https://www.pixiv.net/en/users/14347466">Harusame</a>.How to use Trained on a dataset of the artists Elona art tags that work well are Japanese RPG ones - armor, kemonomimi, cloaks and so on. Many of the images contained more than one character and since it wasn't possible to split everyone up you'll have to use slightly more precise prompting than usual for best results.MISC Trained on Pony SDXL at 1024 resolution with a dataset of 181images. All sample images are raw txt2img upscales, no inpainting or helper embeddings used, e.g - your outputs should look better with more than minimum effort. The training was done on a small subset of the artists images, as such the output of this LoRA is not an accurate representation neither in quality or style of the source material. Go support the <a rel="ugc" href="https://www.pixiv.net/en/users/14347466">artist</a> if you want the good stuff.

Files with SHA256: 1539b920ae9a6a66131076f4d38b7c2fe64dcf991c35dcd6f7133e369d0d8079