Some info
about
jpeg thing Well let me tell you about it.
👩✈️
1. paste A JPEG image into the right area.
Do this by clicking in the square once and then pressing ctrl+v or cmd+v.
2. the image should now be loaded.
to clear it press
Clear Image in the control section (top left). Now you can paste another image.
3. at any point,
"save" the image by pressing Kopimi image to clipboard
in the control section. Now paste it in an image editor or whatever. Tumblr new post dialog.
📵
4. to mess with the image, look at the bottom left section.
it should be filled with numbered segments. these are the different parts of the jpeg image. they can be manipulated in various ways:
U/D. up/down.
these just reorder the segments in the image. Different segments have different functions and
sometimes their order matter.
Number Input.
This shifts the segment data around, but leaves some important bytes in place (at least in some types of segment).
The number in the black box to the right of it is the length of the segment. Shifting by this amount gives you the same as zero shift.
Segment-Specific Buttons.
Sometimes there are extra buttons for some segments.
R usually means randomize.
H also means randomize.
6️⃣
TIP.
segments to the bottom of the data are usually safer to mess with. try moving a single huffman table from bottom to top, or vice versa.
TIP.
it can be beneficial to have the browser console open to quickly notice when the browser consider the image corrupted. it's very easy to "brick" the image by not figuring out how to un-corrupt it.
TIP.
it's not necessarily the best thing to retreat at the first corruption though, especially when rotating table offsets... sometimes there are "islands" of working values scattered among those who corrupt.
TIP.
"auto" bruteforces working values for some offsets.
It only works in chrome (sorry), and is very "full speed" in its approach.
It's not too fun eitherway i think; it stems from a previous way of going about the JPEGs, when i didn't want to parse them much.
(What only works in chrome is the WebCodecs API that lets me try to decode the generated jpegs and see if they are corrupt (remember, we're bruteforcing here). I tried with good old <image> tags first which "would've" worked in firefox, but that ate 40 gigs of RAM immediately lol. (Before decoding in a webworker it also froze my computer to the point where the trackpad's haptic feedback shut off, something i've never had happen before))
TIP.
there is no undo :) yay
🍘
🈲
no thanks