π¬ MY SETTINGS π¬
πΈ CLIP SKIP: 2
π² CFG Scale: 4.5 | 5 | 5.5 | 6 (Optimal)
π§Ό Samplers / Schedulers: Euler A, DPM++ 2M SDE, SGM Uniform, BETA, Karras
π£ Steps: 20 β 40
β¨οΈ Hi-Res Fix: 4x-UltraSharp or Foolhardy Remacri (Don't use it now)
π Adetailer: Use for Face and Eyes if you have problems with them
β¨ PROMPT PARAMETERS β¨
βοΈ Quality (Illustrious): masterpiece, best quality, very aesthetic
β watermark, signature, copyright name, artist name, logo, text, english text, artist logo, twitter logo, twitter username, character name, patreon name, patreon logo, worst quality, low quality, normal quality, lowres, bad , bad hands, extra digits, missing fingers, error, blurry, text, child, childish, , ,
π‘ ADDITIONAL NOTES π‘
β οΈ Scheduler: You can use your own scheduler, but it is not recommended.
βοΈ CFG: You can adjust Classifier-Free Guidance to a greater or lesser extent, but the settings listed above are those I used.
π VAE: Baked into Model.
Description
𧬠Full Description: πΎπ»πΆπ°πΊππππ
Based on: Special [Whyxeo_Large Model - Fine-Tuned Model x Noob EPS version x RouWei]
What's new in this version:
π₯ Aesthetic details: The purpose of this model is aesthetics, deeper tones, work with effects, exposure, detail reduced for a more artistic style + excellent abstract and portrait images.
π¦ Furry support - Support for Furry characters.
β¨ Better effects: Fine-tuned lighting and visual effects for an artistic look, etc., etc.
π¨ LoRA-friendly: Not full support for style LoRAs, since the model has a slight style mixed in, but experiments are welcome.
π NoobAI Epsilon version 1.1 + RouWei 0.8
π A Personal Note;
Stable version, contains a minimal mixed style, the model remains flexible, works with Danbooru authors and their tags, and It may suit those looking for an aesthetic direction, deeper tones in the color palette. I've worked on more abstract prompts and exposure; otherwise, everything depends on your ideas.