Created using frames from videos created using GANS. Original video with real humans http://www.nuum.co/1920×1080/
Unable to average the frames, I ended up drawing the frames on top of each other with very high transparency. Mathematically this should yield a similar result? If transparency level == number of frames.
I’ve worked with a similar technique in the past trying to replicate a pinhole camera image i.e. creating a digital pinhole (long exposure) image with a video.
There are digital artefacts that end up looking like paper texture in what is supposedly empty spaces (the background).