使用模型收缩，SAG和FREE U作为生成图像的模型微调，使用第二个ksampler接收放大的letent以达到upscale的效果，并且增加facedetail用于让面部美丽。Model shrinkage is used, SAG and FREE U are used as the model for fine-tuning the generated image, a second ksampler is used to receive the magnified detail to achieve the upscale effect, and face detail is added to make the face more beautiful.流程较为简单，而且几乎没有使用什么第三方节点（可能除了impactpack，facedetailed相关的功能需要用到它），模型微调引入了一些随机性，让他在生成较小图像（512*512）的时候也能有很好的效果The process is relatively simple, and it uses almost no third-party nodes (except for impactpack, which is needed for facedetailed related functions). The model fine-tuning introduces some randomness, allowing it to perform well even when generating smaller images (512*512).

shrinkupscaleandadetail t2i workflow

可以使用不同的模型用作facedetailed，都可以尝试

Files with SHA256: 3b9fa70588be9542e0499518f704076f3321cbb61fa91ef830540bb328436436