Hello everyone,

New IL Version Description:

Well, well, here we are again, on my way to find a new sweet spot for my IL model while creating some new Halloween images. (Spoiler: “looking through a fence at another person” sucks 🤮!!!)

And which model could serve a better purpose for that than my personal waifu — Orihime? 😅🤣 She had a rough time, that much I can tell.

But all jokes aside, why another version?

Simply because I found a better sweet spot for my current LR: 800 steps. It makes her look even closer to her anime appearance while still keeping the model flexible.

So, what about the prompting thing I mentioned?

Well, I tried to generate a picture where the viewer was looking through a fence at her while she was captured by skeletons. I ultimately failed, but I found a (more or less) stable alternative: cages.

Why? Well, there are useful keywords like “in cage” and “behind bars”, which make it much more likely to generate something decent.

And here, Anima really gets a big point, because while IL struggles to generate such complex scenes, Anima managed to get it right instantly!

_____________________________________________________________________

(older) Anima Disription

these are models of my favorite girl from Bleach, Orihime Inoue.

12/09/2026 — Added Anima version 1.1 with many changes (see: Changelog)

"Yet another version of her??? Why???"

Sorry, but hear me out: I actually used her to greatly improve my Anima workflow. I figured out a lot of things with the help of AI and plenty of testing.

The most valuable thing I learned is that Anima produces much better results with natural language prompts (which the Anima checkpoint page actually tells you). The thing is, natural language prompting isn't really my thing, so I asked AI to give me a general quality prompt for generating anime images:

A polished high-end anime illustration with refined layered shading and clean, elegant linework. The image has strong material separation, visible midtones, subtle texture, delicate folds and controlled highlights. Hair is rendered with fine individual strands and soft highlights, while the eyes are luminous, detailed and expressive. Lighting is balanced and cinematic, with soft directional light, restrained bloom, subtle rim light and clear tonal contrast. The scene contains fine atmospheric details and depth without becoming overly busy, with crisp silhouettes and a highly finished professional anime illustration look.

The difference? See for yourself:

Left: Danbooru quality prompt (masterpiece, best quality)
Middle: Danbooru quality tags + natural language quality prompt
Right: Danbooru quality tags + natural language quality prompt + anime coloring

"Anime coloring" somewhat makes the image look worse, so why would you use it?

Well, with just the normal quality prompts (Danbooru + natural language), the images can sometimes look almost too polished and a bit too AI-ish. That's why I personally like using anime coloring to tone that down a little. However, it's completely optional.

Prompt: look at "Trigger Words" on the right side 👉

Changelog: look at "About this Version" on the right side 👉

My Settings (forge neo)*:

*note, the example images only use high-res fix.

Checkpoint: waiNSFWIllustrious_v170 for IL, waiANIMA_v10Base10 for anima

Sampler: Euler a for both IL and Anima

Scheduler: Automatic Karras for IL, Normal for Anima

Samplings Steps: 30 for both IL and Anima

Hires.fix: R-ERSGAN 4x+Anime6B, 20 Hires steps for both IL and Anima, Denoising Strength: 0.3-0.5 for IL and 0.2-0.4 for anima

Extensions (nice to have for little more Details):

Adetailer (highly recommended)