案例一:img2img(旧照片人脸修复+高清放大)
操作步骤:
1、大模型选择“beautifulRealistic_brav5.safetensors”或者“chilloutmix_NiPrunedFp32Fix.safetensors”(其他写实模型也可以自行测试效果如何);
2、确认设置中的人脸修复选择为“GFPGAN”,参数设置为1;
3、上传照片到img2img;
4、提示词留空;
5、按照如下参数设置:
img2img设置参数:
Resize mode: Just resize
Sampling method: DPM adaptive
Sampling steps: 20
Restore faces: √
Resize to:
Batch count: 8(根据个人电脑情况而定,电脑配置不够可以保持1,然后多次生成来选择最优图片)
CFG Scale: 7
Denoising strength: 0.05
Seed: -1
6、挑选最优的图片发送到后期处理“Send to extras”
7、按照如下参数设置
extras参数设置:
Scale by
Resize: 2 – 4 (根据自己的需要设置放大倍数)
Upscaler 1: SwinIR_4x
Upscaler 2: None
Upscaler 2 visibility: 0
GFPGAN visibility: 0
CodeFormer visibility: 0
CodeFormer weight (0 = maximum effect, 1 = minimum effect): 0
原创文章,作者:朋远方,如若转载,请注明出处:https://caovan.com/sangejingdiananlixuehuiwannengdetushengtu/.html