data mosh
hello
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"
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 ?
thanks
jjr
fwiw, the "technique" employed in the first example is based on
compression artifacts resulting from use of a compression codec that
uses differencing - i.e. only updating parts of the image that has
changed from the previous frame. for example: if you have a static
image of a room and then a figure walks through the frame, the codec
only updates the part of the image around the figure, leaving the
background alone. if you compress a video with one of these codecs
(there are many, both mac & pc) play it back and then jump to a later
place in the video while playing you will see this in action. a very
cool effect, though somewhat difficult to control.
hope this helps
cheers
b
On Feb 19, 2009, at 12:23 PM, jimmy joe roche wrote:
>
> hello
> 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"
> here is a link
>
> http://www.youtube.com/watch?v=6LG39Wp7OzQ
>
> http://vimeo.com/3256023
>
> I've been trying to figure it out for a while and was wondering if
> it is possible to achieve in MAX ?
> thanks
>
> jjr
>
>
Quote: BaltimoreMAX wrote on Thu, 19 February 2009 09:23
----------------------------------------------------
> hello
> 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"
> here is a link
>
> http://www.youtube.com/watch?v=6LG39Wp7OzQ
>
> http://vimeo.com/3256023
>
> I've been trying to figure it out for a while and was wondering if it is possible to achieve in MAX ?
> thanks
>
> jjr
>
>
----------------------------------------------------
It's a glitch based on a compression algorithm that's not performing as expected. In video, I guess I'd start trying to apply motion jpeg compression at low settings to see what happened.
In audio, you could try the same thing through some lossy compression codec, like mp3. Most of the time the result will be lame, but sometimes.
I think this process is intrinsically non-realtime.
mz
More:
http://createdigitalmotion.com/2009/02/18/data-moshing-the-online-videos-my-god-its-full-of-glitch
2009/2/19 jimmy joe roche
>
> hello
> 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"
> here is a link
>
> http://www.youtube.com/watch?v=6LG39Wp7OzQ
>
> http://vimeo.com/3256023
>
> I've been trying to figure it out for a while and was wondering if it is
> possible to achieve in MAX ?
> thanks
>
> jjr
>
>
>
On Thu, 19 Feb 2009, mzed wrote:
-&e->In audio, you could try the same thing through some lossy compression
-&e->codec, like mp3. Most of the time the result will be lame, but
-&e->sometimes....
Was this pun even intended? (mp3, codec, lame... *snort*)
If you want to play with emulating mp3 degradation, use the same
techniques mp3 encoders do to shave data off audio:
Mp3 is an FFT-based compression scheme. So it's pretty easy to go for the
jugular in getting those kinds of artifacts by just
dropping/averaging/bitcrushing/summing/and even delaying data inside a
pfft (which starts to do some really weird things, as the delayed signals
end up landing in different bins, creating a frequency as opposed to pitch
based transpose, which sounds horrible, which some sick puppy might
actually want). Another thing to try is using [cartopol] and tweaking or
dropping the phase data (preserving just amplitude) and then
[poltocar]'ing it right back.
Back to the original question regarding moshing video rather than munging
audio, Recipie 27:Debris in the Jitter Recipies:Book 3 does a good job of
starting you off with that kind of effect, except in a way more
controllable fashion and possible in realtime:
https://cycling74.com/story/2006/2/17/153427/406
andyw
As i understand it, it's a matter of removing the frames that serve as keyframes. Once the keyframes are gone, there's no reference for the codec, so it just goes along and does what codecs do, which is recalculate the image based on assumptions. However, by removing the keyframe it's assumptions are now "wrong" and you get the lovely effect.
So, now you ask, how do you know which frames are keyframes? But this article goes inside the process used in Kanye's video in good detail:
http://www.existingvisual.com/motionographer/2009/02/20/tintori-and-nabil-breaking-your-internets/
An algorithmic approach might be to analyze an MP4 videostream for cuts (where the difference between 2 frames is relatively large), then cut out the surrounding frames. You'll probably knock out the keyframe in doing that, and the next time you play the video, it will be screwed up! If you wanted to do this realtime, you'd have to have some amount of buffer or dual head playback so you could look ahead for a keyframe, then knock it out before you play it back to the viewer.
Quote: AndyW wrote on Thu, 19 February 2009 11:42
----------------------------------------------------
> On Thu, 19 Feb 2009, mzed wrote:
> -&e->In audio, you could try the same thing through some lossy compression
> -&e->codec, like mp3. Most of the time the result will be lame, but
> -&e->sometimes....
>
> Was this pun even intended? (mp3, codec, lame... *snort*)
>
All puns are intended.
;-)
mz
I made a patc h that did this. All you do is export the video as sorenson 3 with 0 keyframes in the compression setting. then make a max looper patch and you can loop and blur and smear the heck out of it. Its really fun.. you get very painterly images sometimes... its really trippy and messed up.. this is probably more for the jitter board though
Heya, I made a tutorial on my methods for achieving the effect in the Chairlift vid:
But it is a great relief/regret to see that it is easily possible in Jitter. One of these days I'm just going to have to buy the program and officially solidify my future as being on my computer every second of the day.
Oh, also, there is apparently a way to datamosh live (viz. Ramsey Arnaoot's Audio-Visual Sampling Synthesizer), due to the way that VLC plays wmv's. Apparently you can scrub or skip between clips and it will datamosh right before your eyes, especially if you strip the footage of keyframes. Of course, this is hard to control, but I believe it's how artists like Takeshi Murata and Paul B. Davis do some of their work.
Ive been datamoshing live for about 3 years with a few jitter patches Ive made, Its not hard to do, and now andrew has made a glsl shader that basically does this. I had also made some things with the cv objects: check them out on our vimeos:
http://www.vimeo.com/3116106
andrews shader
http://www.vimeo.com/3785971
my patch with jit.cv.hsflow+repos
note that these can produce looks from realistic datamosh to watercolor paintings.
I smell the new visual autotune.
now please, someone erase that song from my brain.
ugh, dont even go there.
My friend jimmy edgar used to use autotune on all his tracks back in the day. It was pretty funny. I hate it when I do something and then it becomes trendy, it makes me not want to do it. I didn't even know the term datamosh, I have just been messing with video compression since I noticed it did cool things back in college. I feel like the whole cool aspect of it was that nobody was doing it. It does bother me a bit the idea of having a tutorial on doing something that is cool mainly because of its obscurity, and thus making it commonplace. I watched the videos and they were entertaining, but that aspect of "man people wouldnt even be looking for this stuff if it werent for konye" I wish that people had just kept this as a secret tactic for embellisments, like what David Orilly is doing, he used it to great effect in his video for Vsnares, and his compression reel is great for its sense of humor.
Sorry, just venting... not trying to hate on "datamosher", seems like a cool guy, I just kinda wish these didn't exist hehe
hey AndyW
"Mp3 is an FFT-based compression scheme. So it's pretty easy to go for the
jugular in getting those kinds of artifacts by just
dropping/averaging/bitcrushing/summing/and even delaying data inside a
pfft (which starts to do some really weird things, as the delayed signals
end up landing in different bins, creating a frequency as opposed to pitch
based transpose, which sounds horrible, which some sick puppy might
actually want). Another thing to try is using [cartopol] and tweaking or
dropping the phase data (preserving just amplitude) and then
[poltocar]'ing it right back."
i like it, do you have example of patchs to share?
dank
freaka