I have enjoyed using the plate solving based polar alignment tool. The GUI is very nice, and it is fun to use. I have used it a lot.

I recently tested the accuracy of this procedure and found that it can be 5 arc minutes off no matter how well I try to perform the procedure. I can get better accuracy manually using the FITS viewer. Here is how it works:

1) Start out with an OTA and mount position where the RA rotation axis contained in the FOV.
2) Rotate the mount while taking a 20 second sub at high gain, preferably at low resolution (4x4 binning) for maximum processing speed.
3) After verifying that the center of the star tracks is in the FOV, save the image as circles.png.
4) Load circles.png back into the FITS viewer. It opens it up in a new tab, say, tab 2. You can switch between tabs 1 and 2 using Ctrl-Tab.
5) Start taking 1 second subs at high gain in a loop (the cycle button). Tab 1 will now refresh with a new image every second.
6) Go to the circles tab (tab 2) and put your cursor on the center of the tracks.
7) Switch back to tab 1 without moving the cursor (use Ctrl-Tab).
8) Double-click and a bull's eye is created. This is now exactly where the RA axis is located in the FOV.
9) Use the Alt/Az controls to move the NCP on top of the bull's eye. This is easy because the subs refresh every second.

You can eyeball the position of the NCP in step (9) by looking for the below asterism:

The eyeballing accuracy is about 1', which is better than the 5' error that I got with the plate solving based polar alignment. This is apparently not unusual as it was also noticed by some users who tested similar tools with similar results. For reference, the star tracks corresponding with this image are here:

Step (1) is not trivial if you start out from scratch. However, if you have the habit of putting the scope on marked feet locations on pavers, for instance, you will have little trouble doing this. If it is a problem, you can start out with a regular plate solving based polar alignment and finish up with the above for best results.

This method is a derivative of my old method where I used a DSLR and a piece of tape with a needle hole in it to mark the center of rotation on the DSLR's LCD screen. Then I would take subs to move the NCP to the needle hole. No computer required, and quite accurate.

Hope this helps. It is easy to use and only requires one rotation, so it is relatively quick. It is fun, too. I am curious to hear what kind of accuracy others achieve with the plate solving based method.