I found a new AI tool Shakker, a best image to image tool. You can try it via https://www.shakker.ai
这是一个针对交通工具设计而优化的模型,基于SD1-5微调,并融合了部分DreamShaper的权重.做这个模型的本意是用来做汽车设计,结果发现此模型在其他方面效果也意外的好.模型整体风格偏向一种优化的摄影感,只需要简短的提示词就可以生成非常漂亮的光影.希望此模型能够帮助到各位,尤其是在交通工具设计方面.
This is a model optimized for vehicle design, fine-tuned based on the SD1-5 , and incorporating partial weights from DreamShaper. The purpose of making this model was to do car design, but I found that the model was surprisingly good in other aspects. The overall style of the model is biased to an optimized sense of photography, requiring only short prompt words to generate very beautiful light and shadow. I hope this model will help you out, especially in vehicle design.
Generation data: CFG=7, steps=20, sampler=Euler a, seed=1986
以上链接是一些sd模型的测试图对比,为了对比模型的基础能力,只使用了提示词,未使用lora等额外节点. 除了本模型外还包括dreamshaper/sd 1-5/ sd xl 1.0,此页面会不断扩充.
The above link is the test map comparison of some sd models. In order to compare the basic ability of the model, only the prompt words are used, and additional nodes such as lora are not used. this page includes models as dreamshape r/sd 1-5 / sd xl 1.0....... , this page will be updating constantly.
提示词说明
1 car exterior 生成外饰
1 car interior 生成内饰
bright background 浅色背景
aesthetic background 美学背景
dark background 深色背景
indoor background 室内背景
outdoor background 户外背景
front view 生成前视图
side view 生成侧视图
rear view 生成后视图
top view 生成顶视图
front 45 degree view 生成前 45 度视角
rear 45 degree view 生成后 45 度视角
bird view 从较高的视角拍摄
driver view 控制内饰角度为主驾侧
passenger view 控制内饰角度为副驾侧
middle view 控制内饰角度为主副驾中间
year 2020 可通过 year xxxx 来控制生成的年代感,例如输入 year 1960 会使生成的图像具有 60 年代的设计特征,也通过输入 year 2030,year 1960 使生成的图像同时具有未来 和复古的感觉