diff --git a/README.md b/README.md
index 1c05dde9e038b3dad339abddf0bb7479fcc0d3d9..baa0ebbdf27a1c567a3c0de60d509cf2e652f629 100644
--- a/README.md
+++ b/README.md
@@ -44,7 +44,7 @@ plt.plot(data[:,0],data[:,1],'kx')
 plt.title('data')
 ```
 
-![randomly generated mixture model data](http://www.mit.edu/~mattjj/github/pybasicbayes/datapoints.png)
+![randomly generated mixture model data](https://raw.githubusercontent.com/mattjj/pybasicbayes/master/images/data.png)
 
 Imagine we loaded these data from some measurements file and we wanted to fit a
 mixture model to it. We can create a new `Mixture` and run inference to get a
@@ -95,7 +95,7 @@ for scores in allscores:
 plt.title('model vlb scores vs iteration')
 ```
 
-![model vlb scores vs iteration](http://www.mit.edu/~mattjj/github/pybasicbayes/scores.png)
+![model vlb scores vs iteration](https://raw.githubusercontent.com/mattjj/pybasicbayes/master/images/model-vlb-vs-iteration.png)
 
 And show the point estimate of the best model by calling the convenient `Mixture.plot()`:
 
@@ -104,7 +104,7 @@ models_and_scores[0][0].plot()
 plt.title('best model')
 ```
 
-![best fit model and data](http://www.mit.edu/~mattjj/github/pybasicbayes/fit.png)
+![best fit model and data](https://raw.githubusercontent.com/mattjj/pybasicbayes/master/images/best-model.png)
 
 Since these are Bayesian methods, we have much more than just a point estimate
 for plotting: we have fit entire distributions, so we can query any confidence
diff --git a/images/best-model.png b/images/best-model.png
new file mode 100644
index 0000000000000000000000000000000000000000..ef4e9b844aac27a4194366354fcbc1833e1df59f
Binary files /dev/null and b/images/best-model.png differ
diff --git a/images/data.png b/images/data.png
new file mode 100644
index 0000000000000000000000000000000000000000..c76e0c002c206fb55d9272ffafaa602ec1e89413
Binary files /dev/null and b/images/data.png differ
diff --git a/images/model-vlb-vs-iteration.png b/images/model-vlb-vs-iteration.png
new file mode 100644
index 0000000000000000000000000000000000000000..3b35ccf56cac19ff695f583041d4ff20ddc4890d
Binary files /dev/null and b/images/model-vlb-vs-iteration.png differ