diff --git a/readme.md b/readme.md
index 21e809c..70c2aaa 100644
--- a/readme.md
+++ b/readme.md
@@ -29,21 +29,21 @@ final result on face swap directly from original big image:
 <p align="center">
     <img src="https://s2.ax1x.com/2019/02/01/k3BCe1.png">
 </p>
+<p align="center">
+	<img src="https://s2.ax1x.com/2019/02/01/k3cQTs.png">
+</p>
+
 
 <p align="center>
 	<img src="https://s2.ax1x.com/2019/02/01/k3cQTs.png">
 </p>
-
-<p align="center>
-	<img src="https://s2.ax1x.com/2019/02/01/k3cQTs.png">
-</p>
-
-<p align="center>
+<p align="center">
 	<img src="https://s2.ax1x.com/2019/02/01/k3c3Yq.png">
 </p>
 
 
 
+
 As you can see above, we can achieve **high resolution** and seamlessly combination with face transformation. final result on face swap directly from video (to be added soon):