「Wanna come over for tea?」

Redistribution is prohibited.

ABOUT THE MODEL

Hu Tao from Genshin Impact

20260405v1

Trained on NoobAI-XL V pred 1.0

This model has been trained many times, but the results still aren’t very good. Even in the current version, the consistency of the style hasn’t reached my expectations.

The main issue is that a few data in the dataset have a disproportionately large impact on the overall performance of the model, especially some AI-generated data that were added just to introduce additional concepts.

It’s also possible that this character is already very firmly embedded as prior knowledge in the model, and my caption ended up triggering it, which interfered with my training.

I originally planned to add another set of Hanfu outfits, but the results were very unstable, so I eventually removed them all.

Features

• 2+ outfits.

• Anime style.

Issues

• Bad hands — adding “finger ring” to the negative prompt, or removing descriptions of hand accessories, might improve the result.

• The style consistency is not stable enough.

• When the face becomes blurry, consider using ADetailer to fix it.

• When the output becomes blurry, consider using Hire.fx to fix it.

TRIGGER WORDS

Default Outfits

hu7a0,
black hair, long hair, bangs, twintails, black hat, red eyes, black hutao-dress, black shorts, white socks with red ribbon, black loafers

Cherries Snow-Laden

hu7a0,
black hair, long hair, bangs, updo, twintails, white beret, sunglasses on hat, red eyes, white hutao-outfits, red cape, red skirt, black thighhighs, white hutao-boots,

Requirements

Requirement

Checkpoints: NoobAI-XL V-pred 1.0

Recommended Tools: Forge

Sample: Euler

CFG: 3~5.5

Recommended weight: 1~0.6

Hires.fix: Recommended Upscale value is around 1.2~2.0, denoising strength is 0.1~0.2.

Recommended size: 768*1344, 832*1216, 896*1156, 1024*1024, 1344*768, 1216*8

32, 1156*896

I’ve set the gallery moderation of this post to “R”, so explicit content may not appear on this page.