nice. v4.0 is improved details and colors than v3.0

v4.0 default style is too strong, artists prompt are very weak or do nothing.

@ZU_KI i cant create [ankle socks]

nice,WeiDa,no need to say more

I like it, which version is less "cunny" though? I can't make milfs.

V4 seems to be a little over-trained, not nearly as much variety with results anymore.

This is my favorite model now. I look forward to the future of its development.

I'm really enjoying this model, it mixes well with some of the western artists I have been trying lately for some pretty nice new results

After much extensive use of this model, I can say that it really is one of the best anime IL models out there.

Firstly, this model has a strong style to it, including representation of age (youthful) and culture (less racial distinction), which is expected. This results in consistently good results for people who expect them. You wouldn't be here otherwise anyway.

How it does anatomy is amazing out of the box with the right prompts. I've used no Loras with some complex positions, though some very specific ones will need them. The face and hands look good, but need some inpainting support (I use Adetailer).

Very good at NSFW. If you have trouble making it listen to some specific prompts, I recommend using booru tags; look at Danbooru's tag wiki, there may be a specific tag to what you want. I should note that there are some characters (with low posts in boorus) that it doesn't know, but other models do. Doesn't matter personally since this model achieves much prettier results, and it can be fixed with Loras or enough prompting and styling.

The strong consistent style. is easily influenced by style Loras, but certain prompts and artist tags work amazingly enough. Regarding version models, I usually generate a lot of male images, and v3.0 was perfect at it. v4.0 is different from v3.0 where the latter was good at both genders and can generate mature face and body types. v4.0 however is strongly consistent with more younger characters, and especially prefers females more toward males. So for my use case, I prefer v3.0 more. The difference between both models to me is more of a different use case and model rather than an upgrade/improvement to its predecessor.

Overall, I strongly recommend it for immediate great results out of the box! Use proper prompting and booru tags for better results. I personally rank this model above T-ponyai3, Animagine 4, Pony V6, and the current NoobAI model. Maybe it's because the IL models are just better.

Update: v5.0 dropped shortly after writing this (just like v4.0 on my previous comment lmfao) and the tl;dr is the styling of males in that version is closer to an "AI art" feel IMO, and that body musculature is more exaggerated for boys. Not mentioning anything about the body will still give you abs and such, and it will lean more towards bulky bodies.