最近看见一款新的图生图ai工具 shakker！！来 试试看！ https://www.shakker.ai

基于手冢治虫的绘图的训练，不允许用于任何商业用途！请摸着良心，用老爷子不会讨厌的方式使用吧。

使用2d风格的大模型的时候需要设置较低的权重，0.3~0.5应当是比较合适的数值！和写实模型搭配的话0.85比较合适

数值越高画风以及表现手法都更有古早感，会更像手冢治虫的风格、同时画质可能更低。合适数值内的用色以及比例很有韵味，有些许彩铅质感，但具体是哪个方向的画风靠roll。似乎泛用性较好，经试验和角色lora能友好相处，和部分画风古早的角色模型一起用还可以一定程度上还原其老画风，同olfn(Old-Fashioned Diffusion) 结合效果更佳。

似乎很容易出现描述外的多人/多余肢体。很久以前训练的所以是textual，没有更新过。

I found a new AI tool Shakker, a best image to image tool. You can try it via https://www.shakker.ai

Drawing training based on Osamu Tezuka is not allowed for any commercial use! Please be nice and use it in a way not be hate.

When matching a 2d style model, you need to set a lower weight of this, 0.3~0.5 should be a more appropriate value! 0.85 works well with realistic models

The higher the value, the more old the style and the expression, the more like the style of Osamu Tezuka, while the quality of the picture may be lower. The color and proportion of the appropriate value are very appealing, and there is a little color lead texture, but the specific direction of the painting style depends on roll. It seems to have good generality, after testing, lora can get along well with the character, and some of the ancient character models can be used together to restore its Old painting style to a certain extent, and the combination effect with olfn(old-fashioned Diffusion) is better.

It seems easy to have more people/extra limbs than described. It was trained a long time ago so it's a textual, it hasn't been updated.