diff --git a/README.md b/README.md
index 0fa7628197aac7..66ca5ffa1d0804 100644
--- a/README.md
+++ b/README.md
@@ -50,6 +50,8 @@ Elaborating further:
 
 If you use numpy, then you have used Tensors (a.k.a ndarray).
 
+<p align=center><img width="30%" src="docs/source/_static/img/tensor_illustration.png" /></p>
+
 PyTorch provides Tensors that can live either on the CPU or the GPU, and accelerate
 compute by a huge amount.
 
@@ -75,6 +77,8 @@ from several research papers on this topic, as well as current and past work suc
 While this technique is not unique to PyTorch, it's one of the fastest implementations of it to date.
 You get the best of speed and flexibility for your crazy research.
 
+<p align=center><img width="80%" src="docs/source/_static/img/dynamic_graph.gif" /></p>
+
 ### Python first
 
 PyTorch is not a Python binding into a monolothic C++ framework.
diff --git a/docs/source/_static/img/dynamic_graph.gif b/docs/source/_static/img/dynamic_graph.gif
new file mode 100644
index 00000000000000..b4f17374e03491
Binary files /dev/null and b/docs/source/_static/img/dynamic_graph.gif differ
diff --git a/docs/source/_static/img/tensor_illustration.png b/docs/source/_static/img/tensor_illustration.png
index 6f44a35c7dba9a..b0039c7f3f3ebf 100644
Binary files a/docs/source/_static/img/tensor_illustration.png and b/docs/source/_static/img/tensor_illustration.png differ