🔹 AII-KO Illustrious🔹
This checkpoint was lovingly fine-tuned on the Illustrious-XL base model. It's my possibly only model release, made possible through the support of someone special.
Focus on general anime style, adapting to LoRA styles and using artist tags, in addition to the ability to use prompts. It generally does not have a rigid style and is more adaptive for use.
🎨 Style & Flexibility
For anime-style images, try: anime screencap, anime coloring
For artist-inspired styles, you can check the link provided in next link to explore and experiment.
https://mzmaxam.github.io/Illustruous-Artists/index.html
Example combinations of artist styles will also be available for inspiration.
(no lineart), (artist:mika_pikazo:0.5),(artist:ningen_mame:0.8),(artist:ciloranko:0.5), (artist:ogipote:0.5),(artist:tianliang_duohe_fangdongye:0.7),(artist:wlop:0.9),(artist:jyt:0.9), (artist:vofan:0.5)
📑Recommended Trigger & Prompts
Trigger Words: masterpiece, high quality
Prompt Structure: quality tags > character count > character names > style/artist tags > physical features & clothing > pose & anatomical details > environment/background > composition modifiers
Prompting Tips: Use Danbooru tags liberally.
Negative Prompts:
⚙️Suggested Generation Settings
Sampling Steps: 25 ~ 32
CFG Scale: ~5 (can go as low as 3 for softer renders; raise to ~7 for more detail)
Sampler: Euler a
ClipSkip: 2
Resolution: 1MP - (1024x1536, 1024x1024, 768x1536...and more)
LORAs trained in this model:
748cm - A variation on the original style
AII-KO Anime Style - Anime style that I really like and I trained in this model.
----------------------------------------------------------------------------------------------------------
In any case, if you want to contact me, please do so via Discord:
Execution model exclusively in TensorArt