Compression Glitch Pixel Selfies

Since the last post I’ve been trying out more ways of batch processing the 1500 face-on selfies using ImageMagick and other tools. This post, and the next, is a documentation of those experiments. If I had to guess I’d say I’m about 10% of the way to something worthwhile.

Through the post I’m going to use the first 9 selfies in my folder, because that’s easier. Here they are as they came from Instagram with no processing.

To recap, when last we spoke I was crunching the photos down to 10px squares and blowing them back up again, then superimposing the original on this fuzzy grid to produce something that wouldn’t look out of place in a 70s fashion magazine.

With that done and posted I remembered a script Antonio had sent me which he was using to crunch photos of the night sky into noise (for the VOID show). It takes an image and saves it 100 times, decreasing the jpeg quality each time, then does this four more times. You get a “sitting in a room” type effect where the original is reduced to static.

Using Antonio’s script as was gave me the sort of result you’d expect from Antonio which didn’t fit into what I was trying for - something transformative and different which respected and referenced the original works. But by editing some of the parameters (mostly reducing the number of iterations) I was able to leverage this effect in a small way to start emphasising patterns.

Here’s the first run, saving the enlarged pixel layer 25 times over, the first at 25% quality, the last at 1% quality.

Note how the borders between the large squares take on this bevelled appearance. That’s quite nice and crops up wherever there’s a horizontal or vertical line.

These have been saved 50 times, the first at 50%, the last at 1%.

This is the same process but running on the original files.

These were saves 100 times, running from 10% to 1% 10 times for each photo. (Number 6 got lost somehow.)

Those were okay but weren’t leading anywhere, so I went back to the compressed squares and started layering the uncompressed squares AND originals over them. Here’s the final result.

I’m pretty pleased with these. The aesthetic is still a little cheesy but it’s heading towards what I want - something that abstracts while emphasising the image. What I particularly like is the pseudo edge-detection that is going on around the faces and the eyes.

Here’s all 1500 selfies in a video.