V2.0 - A slightly cleaner node setup with more refined notes, additional prompt templates to go along with my wildcards - <a target="_blank" rel="ugc" href="https://civarchive.com/models/1885843/kyocel-character-wildcards">https://civarchive.com/models/1885843/kyocel-character-wildcards</a> - and a few new LoRAs. The contrast controller LoRA by reakaakasky: <a target="_blank" rel="ugc" href="https://civarchive.com/models/1523055/contrast-controller-ilnai">https://civarchive.com/models/1523055/contrast-controller-ilnai</a>Can be useful for any generation if you find even after the detailer and color your images are still a smidge flat, or maybe you want to go for that higher contrast look! The Stabilizer LoRA by reakaakasky: <a target="_blank" rel="ugc" href="https://civarchive.com/models/971952/stabilizer-ilnai">https://civarchive.com/models/971952/stabilizer-ilnai</a> I feel this LoRA has a lot more potential than how I use it here, but when generating existing characters with danbooru tags, it can help reintroduce a little bit of that cinematic lighting and detail. V1.0 - After several months of work and a lot of trial and error - I have finally achieved the style I was looking for. I wanted hyper-detailed cel shading, little to no gradient shading, vibrant color, and decent-looking eyes and hands. I think I have succeeded. This workflow uses xquizitSDXLIllustrious_V3 as its backbone, along with some particular prompting, upscaling, and inpainting to achieve the final look.Aside from style, I think this is a fairly decent "complete" workflow for any model you may prefer. It includes text to image, wildcard text prompting, image to image upscale (tiled and non-tiled), as well as inpainting and metadata extraction. SFW images are more often closer to the "Kyocel" style, while NSFW images might take a few extra iterations to get rid of any gradient shading. I have left as many notes inside the workflow as I could think of to help one achieve similar results. In the grand scheme of things, I am a total noob, so any feedback or suggestions are much appreciated.Links to all models and LoRAs can be found in the notes inside the workflow. I hope to create a LoRA out of this eventually that will work with multiple models, but am severely limited by my MacOS hardware, minimal BUZZ, and inexperience with cloud-based GPU clusters. One can dream.I will continue to upload new versions as I refine this workflow to make it even more convenient and easy to use. Any questions - comment or shoot me a DM any time!

KYOCEL MAX - Illustrious Anime Style

Few new additional LoRAs, refined notes, cleaner layout

Files with SHA256: 7d788d828435b51ccbeeca7d51c482ef431314156b9d8798be2e92f2834714da