This Illustrious model creates art in the style of Tatara. They used to have a <a target="_blank" rel="ugc" href="www.alpha-net.ne.jp/users2/tataran/">website here</a> but it appears to be gone. They may still have a Pixiv account.This LoRA is trained on most of the available non-oekaki art, including some sketches.Trigger word is <code>tatara</code>Works with a wide variety of captioning.It knows the following specific characters: <code>syarii woman, oinarisan woman, usiyo woman</code>Recommended negatives: <code>sketch, monochrome</code> (unless you want these of course)Either <code>huge breasts</code> or <code>gigantic breasts </code>are recommended if you want a good resemblance to Tatara's work.The original artwork was fairly low resolution by modern standards, and this may be reflected in this model which is unregularized.While evaluating Illustrious, I realized this dataset (which I <a rel="ugc" href="https://civarchive.com/models/480247">trained on Pony</a> earlier) would perform way better on Illustrious, and lo and behold, this was indeed the case.

Tatara Artist Model (Illustrious)

Better training on Illustrious v1.0, with a Min SNR of 5.0, and upscaled training data.

Files with SHA256: 6d74ca3cd3b86efbaf6c381d383b7c41ccc10b345853dc27f3174b60d6ee9c89