Show HN: Vector Video
theblackbox.ca
theblackbox.ca
And they also do upscaling via a high quality sinc filter: https://en.wikipedia.org/wiki/Whittaker%E2%80%93Shannon_inte...
The main issue you have is gif artifacts from the extreme color quantization. Something like this might help: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.422...
I've had the most success with the Farneback algorithm. There's even a nice CLI utility called Butterflow which wraps ffmpeg+OpenCV Farneback, and allows you to interpolate video with a one-liner on the command line: https://github.com/dthpham/butterflow
Here's an example of my results applying Butterflow to satellite images from Himawari - the first half is interpolated, while the second half shows the same frames uninterpolated: https://www.youtube.com/watch?v=8knZ2-cys6M & a compilation of a few more: https://www.youtube.com/watch?v=1lPsA8J3U0Q
Try animating a large cube moving into the logo, and using the boolean subtract modifier between each mesh and this cube. Place the cube on another (invisible) layer, et voila.
You can leave your camera fixed in the same place, too.
EDIT to add: I should have mentioned, I think this is a very cool idea! I might just steal it one day ;)
Steal it by all means!
I basically mean replacing your animated camera with the cube, but still using a [fixed] orthographic camera to look at the face of the solid as the cube cuts it away.
I tried using your scripts but Meshlab on my system really doesn't like the data and refuses to produce any output :(
Basically, I'm talking about some kind of voxel representation that is more or less "scale invariant".