diff --git a/README.md b/README.md
index f33eced..11802b6 100644
--- a/README.md
+++ b/README.md
@@ -1,10 +1,37 @@
-# Ubuntu Howdy
+# Howdy for Ubuntu
 
-Windows Hello™ style authentication for Ubuntu
+Windows Hello™ style authentication for Ubuntu. Use your build in IR emitters and camera in combination with face recognition to prove who you are.
 
-Notes:
+### Installation
 
-tail /var/log/auth.log
+Fist we need to install pam-python, fswebcam and OpenCV from the Ubuntu repositories:
 
-/etc/pam.d/sudo
+```
+sudo apt install libpam-python fswebcam libopencv-dev python-opencv
+```
+
+After that, install the face_recognition python module. There's an excellent step by step guide on how to do this on [its github page](https://github.com/ageitgey/face_recognition#installation).
+
+In the root of your cloned repo is a file called `config_template.py`. Duplicate this file and call it `config.py`. The `device_id` variable in this file is important, make sure it is the IR camera and not your normal webcam.
+
+Now it's time to let Howdy learn your face. The learn.py script will make 3 models of your face and store them as an encoded set in the `models` folder. To run the script, open a terminal, navigate to this repository and run:
+
+```
+python3 learn.py
+```
+
+The script should guide you through the process.
+
+Finally we need to tell PAM that there's a new module installed. Open `/etc/pam.d/sudo` as root (`sudo nano /etc/pam.d/sudo`) and add the following line to the top of the file:
+
+```
 auth sufficient pam_python.so /path/to/pam.py
+```
+
+Replace the final argument with the full path to pam.py in this repository. The `sufficient` control tells PAM that Howdy is enough to authenticate the user, but if it fails we can fall back on more traditional methods.
+
+If nothing went wrong we should be able to run sudo by just showing your face. Open a new terminal and run `sudo -i` to see it in action.
+
+### Troubleshooting
+
+Any errors in the script itself get logged directly into the console and should indicate what went wrong. If authentication still fails but no errors are printed you could take a look at the last lines in `/var/log/auth.log` to see if anything has been reported there.
diff --git a/compair.py b/compair.py
index a48ea27..c48aec5 100644
--- a/compair.py
+++ b/compair.py
@@ -21,7 +21,7 @@ encodings = []
 tries = 0
 
 try:
-	encodings = json.load(open(os.path.dirname(__file__) + "/models/lem.dat"))
+	encodings = json.load(open(os.path.dirname(__file__) + "/models/" + user + ".dat"))
 except FileNotFoundError:
 	stop(10)
 
diff --git a/config_template.py b/config_template.py
index 8b6c65a..9ba9900 100644
--- a/config_template.py
+++ b/config_template.py
@@ -3,7 +3,7 @@
 certainty = 3
 
 # The number of frames to capture and to process before timing out
-frame_count = 120
+frame_count = 80
 
 # The /dev/videoX id to capture frames from
 # On my laptop, video0 is the normal camera and video1 is the IR version
diff --git a/pam.py b/pam.py
index a5b5ba5..16a90cd 100644
--- a/pam.py
+++ b/pam.py
@@ -9,7 +9,7 @@ def doAuth(pamh):
 		print("No face model is known for this user, skiping")
 		return pamh.PAM_SYSTEM_ERR
 	if status == 11:
-		print("Timeout reached, ould not find a known face")
+		print("Timeout reached, could not find a known face")
 		return pamh.PAM_SYSTEM_ERR
 	if status == 0:
 		print("Identified face as " + os.environ.get("USER"))