Makes black and white images similar in style to those by the artist Zunta for the Reincarnation Coliseum <a target="_blank" rel="ugc" href="https://www.mangaupdates.com/series/s5x7u47">manga</a>.I trained this LoRA for personal use to assist in cleaning up images from the raws, not prompt specific characters or act like a full-fledged Zunta style model. I decided to release it still since it came out decent enough at both but use it with the above reservation in mind.How to use Both "greyscale" and "monochrome" was used during training so I strongly recommend using them. Removing them to prompt color will usually result in a noticeable style hit. I found the best results for non B&amp;We outputs is simply leaving those in and then prompt colors anyhow. See the separate image for how that might look. As for quality tags using just "score_5_up" seems to work best, higher tags tend to blend in the Pony XL smoothed faux 3D render look too much so. I generally recommend using a more 2D oriented model like <a target="_blank" rel="ugc" href="https://civarchive.com/models/288584/autismmix-sdxl">AutismMix SDXL</a> over the base variant for this reason. For more specific prompt tips please see the sample images or the uploaded zip with training tags.MISC Trained on Pony XL with a dataset of 47 heavily manually curated 1girl images from chapters 1 through 22 at 1024 resolution. All sample images are raw txt2img upscales, no inpainting or helper embeddings used. 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 buy the <a target="_blank" rel="ugc" href="https://comic-walker.com/detail/KC_001662_S?episodeType=first">source material</a> and support the <a target="_blank" rel="ugc" href="https://skeb.jp/@_zunta">artist</a> if you want the good stuff.

Files with SHA256: 8b1023c8b89e66e2828786a412b973df3b9fe2bdcbe8d4ca18d5ce4daafdc3ca