利用IPAdapter Regional Conditioning的风格迁移和通过遮罩构图，让AI生成的图片更加具有可控性，缺点是要得到构图和细节都完美的图片目前还是需要非常多的尝试，Using the style transfer of IPAdapter Regional Conditioning and the mask composition, the pictures generated by AI are more controllable. The disadvantage is that it still needs a lot of attempts to get perfect pictures with perfect composition and details.<hr />V1.Regional Conditioning 3可以将2个主体与一个背景进行结合，得到一张重新构图的图片，You can combine two subjects with a background to get a recomposed picture,整个工作流加载的模型比较多，对低配电脑不太友好，熟练使用comfyui的可以把提示词提取和面部细化部分单独放到新的工作流中使用，以减少显存占用，The whole workflow loads a lot of models, which is not friendly to low-power computers. Those who are skilled in using comfyui can use the hint word extraction and face refinement separately in the new workflow to reduce the GPU memory occupation<hr />V2.Regional Conditioning 2将1个主体与一个背景进行结合，得到一张重新构图的图片。类似局部重绘，但是通过IPAdapter引入了新的风格，让图片呈现出一些不一样的东西。缺点是IPAdapter目前的通病，就是会让新图片的细节出现缺失。但如果用来构图的话，则还是相当有价值的。Combine a subject with a background to get a recomposed picture. Similar to partial redrawing, but with the introduction of a new style through the IPAdapter, making the image appear something different. The disadvantage is that IPAdapter is currently a common problem, that is, the details of the new picture are missing. But if it is used for composition, it is still quite valuable.

Files with SHA256: 3158e3f9ab98504ebc257dc6c86f35022520b97ed70c4a9820e4e897d380b9ee