I’m happy with the result but much of the process is manual so I may have to spend some time slapping together some automation for the next installment.
From: https://www.torontocitylife.com/sectionb/2022/03/25/2-brush-pass/
I’ve been thinking about producing a more immersive rendition of /sectionb. I’ve also been thinking that producing a “Parapsychological Spy Thriller” via conventional means is not be the correct approach. It needs to be a little more artsy, interpretive, associative. Unfortunately, illustration and animation aren’t really my thing.
Although I can draw some basic proportions and I try to pay attention to composition and colour, I can’t produce the type of visual output that modern artificial intelligence can. But as it happens I also dabble in code so it wasn’t long before I was fucking around with Stable Diffusion and similar software. Unfortunately, if I wanted to use the AI to produce short films the still images it spat out would need to be animated using something like morphing — doable but laborious.
By one propitious circumstance a fairly recent upgrade to Stable Diffusion by Deforum popped up in my search results one day and as soon as I saw a few samples I got giddy. Not only is the animated output of DSD dream-like and trippy, which is very apropos for /sectionb, it also improvises around the periphery of supplied prompts/themes in surprising ways, which is also quite apropos.
Initially I tried adding voice narration but it just didn’t fit so instead I converted the text to subtitles/closed captions, chucked in some original music, and after that the video basically just produced itself.
A month and a half between posts might seem excessive but hear me out! I could’ve just posted another static map image but I’m trying to push the envelope a little bit so I thought, “Why not a video?”
Unfortunately, adding a time component introduces a whole bunch of new complications. First I had to alter my code to produce output based on a temporal sequence rather than a single point in time. After that I had to figure out how to produce composite images so that I could add things like the time/date stamp. Then I had to figure out how to actually encode the video. Following that, the clunky user interface needed to be updated in order to accommodate the new features. Then I realized that the data didn’t include any duration information so I had to figure out how to extrapolate it. And then I had to run the extrapolation routine over the whole database which took weeks, no doubt owing to my weak SQL.
Anyways, I find the first product to be kinda soothing and hypnotic and leaves me thinking about what else I could do with it. So without further adieu here’s 24 hours of all C4S calls over Halloween, each call (red dot) growing larger in diameter the longer it remains active:
Regional transit and brooding days seem to go hand in hand for me. Here’s another example from yesterday: me going to meet family to celebrate a birthday out west. Way out west. The flipbook movie is about 13 minutes long (out of a 40 minute trip to Oakville), so give it some time to load. Maybe this is a good time to call your mum?
... with pictures
... with sounds
... with videos
... with opinions
... that's shorter
... that's longer