I recently watched Ian Hubert's lightsaber tutorial. About midway through, he discusses using a plane with an emission material and a clean plate image (as the color) to animate the lightsaber ignition.
I played around with a couple of different videos I had, and in each case, the plane was clearly visible in the render and did not blend well with the underlying video footage.
I ended up finding another tutorial that shows how to make a phone appear to levitate through the use of a mask and clean plate image. I tried this approach as well, but again the mask was clearly distinct from the rest of the video.
I know some steps can be taken to try and minimize the issue--having footage where the lighting/brightness doesn't vary; limiting the number of frames where the clean plate is shown; using compositor nodes to make the mask less sharp; etc. But it still seems like I'm missing something.
As a last resort, I tried an experiment. Instead of using video footage I had shot, I used screen recording software to capture a video of a simple setup in MS Paint. I created a large blue rectangle with a small, thin yellow rectangle inside it. I used a still image of just the blue rectangle as my clean plate. However, after trying both tutorial methods, I was again left with a noticeably visible plane/mask.
Here is a frame render of the experiment using the mask method (I animated the mask moving to cover up the yellow rectangle, but the mask was distinct throughout). 
Does anyone know what I might be doing wrong? Can anyone replicate the tutorials more or less cleanly? I'm stumped.
Thanks in advance.
