diff --git a/README.md b/README.md
index 8f4a953..3e32005 100644
--- a/README.md
+++ b/README.md
@@ -230,7 +230,9 @@ For this mode, we recommend to use 50 steps and guidance scale between 3 and 5.
 
 Besides, if you write a script to generate image captions from the "guess mode" images, and then use the generated captions as prompts to diffuse again, you will get a SOTA pipeline for fully automatic conditional image generating.
 
-![p](github_page/uc2.png)
+![p](github_page/uc2a.png)
+
+![p](github_page/uc2b.png)
 
 Note that the below example is 768×768.
 
diff --git a/github_page/uc2.png b/github_page/uc2.png
deleted file mode 100644
index b2ad69b..0000000
Binary files a/github_page/uc2.png and /dev/null differ
diff --git a/github_page/uc2a.png b/github_page/uc2a.png
new file mode 100644
index 0000000..6765d6d
Binary files /dev/null and b/github_page/uc2a.png differ
diff --git a/github_page/uc2b.png b/github_page/uc2b.png
new file mode 100644
index 0000000..f7b22a4
Binary files /dev/null and b/github_page/uc2b.png differ
diff --git a/test_imgs/building2.png b/test_imgs/building2.png
new file mode 100644
index 0000000..4985fde
Binary files /dev/null and b/test_imgs/building2.png differ