diff --git a/README.md b/README.md
index 4b5f393..0f825ba 100644
--- a/README.md
+++ b/README.md
@@ -16,6 +16,8 @@
 <sup>1</sup>Hong Kong University of Science and Technology,<sup>2</sup>City University of Hong Kong,
 <sup>3</sup>Microsoft Research Asia, <sup>4</sup>Microsoft Cloud&AI, <sup>5</sup>USC Institute for Creative Technologies
 
+<!-- <img src='assets/sample1.png'/> -->
+
 ## Prerequisites
 
 - Python 3.6+
@@ -64,6 +66,14 @@ python test.py --clip_path ./sample_videos/clips/v32 \
 
 Note that we use 216\*384 images for training, which has aspect ratio of 1:2. During inference, we scale the input to this size and then rescale the output back to the original size.
 
+## Colorization results
+
+<img src='assets/sample1.png'/>
+
+<img src='assets/sample2.png'/>
+
+<img src='assets/sample3.png'/>
+
 <!-- ## TODO
 
 - [ ] Release the training code
diff --git a/assets/sample1.png b/assets/sample1.png
new file mode 100644
index 0000000..1b46e9d
Binary files /dev/null and b/assets/sample1.png differ
diff --git a/assets/sample2.png b/assets/sample2.png
new file mode 100644
index 0000000..b44e545
Binary files /dev/null and b/assets/sample2.png differ
diff --git a/assets/sample3.png b/assets/sample3.png
new file mode 100644
index 0000000..a58e7b4
Binary files /dev/null and b/assets/sample3.png differ