Friday, November 11, 2005

friday cat databending: volume

friday cat bending: part 4. (link to part 3)

big post this week, with 8 bends, and most of the files are pretty big. i wanted to try something a little different, so i decided to bend BMP files (a simple format, where i normally work with volatile formats). specifically i wanted to experiment with running various audio effects on the BMP files. others have done this, but the one site i knew of that had a decent catalog of results has since gone dead.

after playing around with a few different fx (which will likely be posted next week), i wanted to experiment with fx that were "reversible": namely, fx that i could run and then rerun with the oppposite settings, such that the end result would theoretically be pretty close to the original. only it wouldn't be quite the same: little digital artifacts would creep in, and these artifacts were what i wanted to explore.

first i tried it with pitch shifting. but pitch shifting in one direction caused very extreme distortion (basically ending up with visual noise), and then shifting it back would break the file. so i gave up on pitch shifting and tried adjusting the volume.

so i took the file leland_poses and reduced the volume ever so slightly, by .09 decibels (reducing it to 98.97% of original volume). even this "minor" change had a pretty significant effect on the pixel data, as you can see in leland_poses-vol9897. then i cranked the volume back up .09 decibels, returning it to its original volume. this resulted in an image pretty close to the original, but with a bit of visual noise, as can be seen in leland_poses-vol-9897-10104.

leland_poses.JPG: original image:

open:this window|open:new window

leland_poses-vol9897.jpg: original -.09db (98.97%):

open:this window|open:new window

leland_poses-vol9897-10104.jpg: previous +.09fb (101.04%):

open:this window|open:new window

i repeated the process, this time increasing the volume by .09db first, then decreasing the volume. the "increased" version is shown here; the increased-decreased version didn't look that different from leland_poses-vol9897-10104, so it is not included for space reasons:

leland_poses-volvol-10104.jpg: previous +.09db (101.04%):

open:this window|open:new window

clearly i wasn't doing enough, so next i increased the volume by 3.73db, then decreased it again by the same. this gave me a suitably distorted image:

leland_poses-vol-373db.jpg: prev. -.09db, then +3.73db:

open:this window|open:new window

leland_poses-volvol-373db.jpg: prev. -3.73db:

open:this window|open:new window

so far, so good. so i went buck wild and decreased the volume by 10.57db, then increased it back. and though decreasing the volume by that much did make it look quite different (see below), after i "re-increased" the volume, it didn't really look much different from leland_poses-volvol-373db above. so the re-increased version is not included below:

leland_poses-vol1057db.jpg: prev -10.57db:

open:this window|open:new window

clearly, the process of decreasing the file's volume then increasing it by the same amount was not corrupting the image very much. in order to cause real change in the image, it would be necessary to increase volume, and to increase it enough that it would introduce distortion and/or clipping into the audio. this distortion would then still be there once the volume was reduced back to normal. so i took that version, increased its volume by 10.57db, then re-decreased it by 10.57db, resulting in the very distorted image we see in leland_poses-volvol-1057db.

leland_poses-volplus1057db.jpg: prev. +10.57db, then +10.57db:

open:this window|open:new window

leland_poses-volvol1057db.jpg: prev. -10.57db

open:this window|open:new window

No comments: