Finally managed a reworked version I'm satisfied with. Since I don't own a very powerful computer, I decided to focus on merging an anime-style model with a painterly aesthetic that can generate good images at lower steps in order to save my PixAI credits and Lora slots. Ideally, I would have liked to make a DiT model, but the styles I wanted are more scarce in that architecture and the generation costs are generally higher than XL generation with similar settings, so I sticked with XL.
Most of my published images have a HiRes Upscale applied to them after generation. Because I want my personal gallery to be as highly curated as possible, my personal workflow consists of using this model to generate iterations for cheap and then apply a hires pass to selected images to upscale and add details, so the generation details on these images will likely not match the recommended steps. The image samples on this model page, however, were published without Highres to serve as a better example of the model
Base Model: Illustrious XL
Recommended Sampling: Euler a
Recommended Steps: 8
Recommended CFG: 2-3
Recommended Quality Tags
Masterpiece, best quality, absurdres, ultra-detailed, very aesthetic, detailed illustration, anime style, (detailed eyes), detailed background, painted background, painterly, texture, rich details, highly detailed, intricate details, intricate,
Recommended Negatives
bad quality, worst quality, worst detail, bad, low quality, bad anatomy, polydactyly, extra digits, extra legs, bad hands, bad hand, bad fingers, fused fingers, missing finger, split fingers, fused arms, fused hands, text, signature, trademark, copyright, logo, patreon,
