data moshing ?
(I originally posted this in the max forum but some one suggested I re-post it here, because its more of a jitter question ((I think)))
have any of you seen these video’s that employ a form of data manipulation, I’ve heard it called "Compression Hack Fade" and "Data Moshing". It has something to do with key frames in compressed files
here is a link
I’ve been trying to figure it out for a while and was wondering if it is possible to achieve in MAX ?
That looks well awesome in hi-res.
Sorry I can’t help you out(just starting).
yes indeed, thats nesa at its best.
while the topic as been dissected around the web, see kottke.org and creativedigitalsomthing. this is a great example of how it can be simulated intentionally in jitter. not an easy feat.
there’s a kind of overview here: http://rosa-menkman.blogspot.com/search/label/artifact
check out the 3 tutorial videos, there’s a link to a datamoshing kit in the 1st
oh man this is fucked up I just made something like this, as did andrew, check our respective vimeos
one of my experiments:
and andrews epicly awesome fluidy goo shader based on the same sort of methods:
Its basically reverse engineering what compression does, which is take a set of pixels and push them in the directions they would be moving, that way they only have to store the movement info and keyframe info rather then every literal frame, this is what most compression is based on these days. If you export a video to sorenson with 0 keyframes you get this awesome effect, and if you use something like jit.cv.hsflow you can input that into jit.repos and use it to push the pixels.
Axiom these are rad,
I’m not sure how your running your camera through compression though? are you using some kind of buffer? how do you patch to get real time zero key frames. can you create a zero key frame file within max? Yes andrews work is mind bending. In your video piece are you changing the way the glitch bleed responds to movement in real time? Are you altering parameters of the effect in real time?
I feel t i n y .
as I said its not really compression, its emulating what happens when compression glitches, which is to take a frame and push the pixels of that frame based on the movement of the following frame, thus with the jit.cv object for flow detection, it outputs the ammount that each pixel moved and with jit.repos, it pushes the pixels by that ammount, then with some experimentation of feedback you can get some really wild results.
Check andrews new sharing post for his flow distortions, then you have all you need to start doing crazy data moshing in realtime!!!
A colleague and I are currently working on trying to develop a patch that is almost "real time" yet stays true to the datamosh aesthetic. Our initial findings were that if you work with certain types of quicktime .mov to .wmv compression, you can literally scratch in some interesting footage via VLC in real time using the playhead like a dj. Now the question is how to integrate that quick writing/reading process and play into the jitter architecture. I’m sure i’ll share some of that with you at Expo74, but if anyone has ideas in that realm, it could be fun.
I dont know if thats possible simply because you wold ahve to write a file in order to compress it, and reopen it in order to mangle it, ths its not really realtime, it wold be fast thogh, jst set up jitter to record into sorenson qt with 0 keyframes and have it automatically open the same file after its done saving it, yo could do a few seconds at a time.
sorry carple tnnel messing with my typgin
I had made a while back a datamoshing patch that just loaded sorenson files with 0 keyframes and did a basic stutter with a range slider, and it would make beautiful painterly imagery.
Axiom-Crux, how do you set it up to export and then re-open, and how do you specify the duration of the file it exports? Can you just do it 30 frames at a time, or something like that?