针对SD画不好武术做的强化训练，例图为用2次元画底图在用写实模型来强化细节（不会的可以去看我的B站教程）直出图可以看我返图区例图，直接用写实可能构图没有2次元好采样器：画底图推荐DDIM，跑图随意。步数推荐33-55，高清步数推荐22-33，V1版这个LORA提供武术动作和烟雾背景，V2版飞天，主要提供轻功的动作构图，V3版偏插画风，主要提供武术动作和插画风构图提示词看例图，V2触发词：Qing Kung（非必要）V3触发词： martial（非必要）<a target="_blank" rel="ugc" href="https://space.bilibili.com/37459046">绪儿已成精的个人空间_哔哩哔哩_bilibili</a>I found a new AI tool Shakker, a best image to image tool. You can try it via <img />,it can help you: -Remix: Upload a picture. Just switch the prompts, and you can create stunning images in the same style. -Style Transfer: Shakker not only extracts the style,but also switches among various styles. Besides, Shakker also offers Object Control,Composition Control,Collage Redrawing etc.

绪儿-武侠 xuer martial arts

This is an integration of my previous 1.5 version of the Wuxia series, Chinese sword series, light Gong.At the same time, the XL integrated version of some Samurai girl content was added.

绪儿XL 武侠.safetensors

v5.0XL

v4.0

v3.0

v2.0

v1.0

xuer martial arts

Intensive training for SD who is not good at drawing martial arts,

The example picture shows a 2-dimensional base map and a realistic model to enhance the details (if you don’t know how, you can check out my B-site tutorial)

If you want to export the picture directly, you can see the example picture in my return area.

If you use realism directly, you may not have a good 2D sampler for composition: DDIM is recommended for drawing the base map, and you can run the map as you like.

The recommended number of steps is 33-55, and the recommended number of high-definition steps is 22-33.

The V1 version of this LORA provides martial arts actions and smoke backgrounds,

The V2 version of Feitian mainly provides action compositions of Qinggong.

The V3 version is more illustration-style, mainly providing martial arts actions and illustration-style compositions.

See the example picture for prompt words, V2 trigger word: Qing Kung (not necessary)

V3 trigger word: martial (not necessary

Files with SHA256: b358927f587eb881f7309bac38c3b492a7959f11bf37170af2ad9841426c28e1