The illusions in this video are pretty cool.

 

The plimsole seemed less convincing than the tape, which in turn was less than the cube - quality of the illusion, ease of shape, or the brain just getting used to being fooled?

I think it's the aspect ratio being so pronounced. The disparity of the actual shape vs its perceived shape grows continually smaller.

 

I printed out the cube. It works better if you close one eye and lose your depth perception, and I think it works well on camera for that same reason.

The close-up of the cube gives a tilt-shift effect (the focus along a horizontal line of pixels is the same regardless of the illusive 3d content of the scene). I think the brain accepts it as a camera effect and ignores it rather than trying to process actual depths from the focus. If you pause the video around 0:26 and stare at it blankly, the actual shape of the paper starts coming through a bit, as the depth cues don't match the perceived cube shape. The sides of the cube seem to bend outward toward the top. The top "near" corner seems to be pushed backward.

 

I think it works well on camera because the brain understands and handles camera effects, and doesn't strain to process them. Also, the camera carefully avoids horizontal movement, which would ruin the effect.

Agree with both above - I also printed out the cube; closing one eye suddenly made it better, and a good photo on an iphone was completely convincing (the little anomalies get blamed on the photo not the subject)