Update README.md

README.md

@@ -12,8 +12,9 @@ inference: true
 Replicant is built by merging several models with fine-tuning WD1.4 and photorealistic SD2.0 models that works with danbooru tags.As with SD1.x, merging individually trained models is better quality than training many concepts at once.This model is a workflow test and is not good enough.  WD1.4 seems to vary greatly in quality with/without Hires. fix.In Replicant, the difference in quality is more noticeable because of the detailed drawings.So I recommend enabling Hires.fix for use.
   
 # Example
-Denoising strength 0.6 is a bit large. Please adjust it.  
+Denoising strength 0.6 is a bit large. I like 0.57 better. 
 The optimal CFG Scale value should also be examined.  
+Hands often multiply. When this happens, increase the value of "extra hands".
   
  ![sample1](https://huggingface.co/gsdf/Replicant/resolve/main/sample_01.png)
 ((masterpiece, best quality)), 1girl, flower, solo, dress, holding, sky, cloud, hat, outdoors, bangs, bouquet, rose, expressionless, blush, pink hair, flower field, red flower, pink eyes, white dress, looking at viewer, midium hair, holding flower, small breasts, red rose, holding bouquet, sun hat, white headwear, depth of field