Well that’s…. pretty impressive Lloura! Think I’ve got a long way to go on Cinema4D 😉
Nuts! Done with deep neural networks apparently: I guess the same sort of thing all those trippy Deep Dream things used that were going around a while ago.
So I’ve seen a few of these things going around recently. Wait for the first cut in the above f*** awful pop song!
Basically: take a few keyframes out of a video stream and let the codec go nuts, overlaying the motion data on the wrong things. Or to be precise, discard some I-frames and keep the P-frames & B-frames. P & B frames store only the differences between their neighbouring frames, while I-frames are “whole pictures”. Delete some of those and you get that whole “pour paint over everything”, “I’ve done a digital sick”, “Im in a really budget version of the matrix” look.
Anyway, it appears this technique now has a name – DataMoshing? There’s even a whole site dedicated to it here. I remember opening JPEGs in text editors to do this kind of thing when I was but a wee slip of a lad, but I thought it was just called Being Fucking Bored 😉 Good to know how far it’s come. Right, I’m off to get AviDemux and have a play!