Has anyone tried a long exposure to see if the motion smears into something discernible? Obviously harder to expose a bright screen without some ND since the shutter speed is the phone's main exposure control
Another idea I had with this concept is to make an LLM-proof captcha. Maybe humans can detect the characters in the 'motion' itself, which could be unique to us?
- The captcha would be generated like this on a headless browser, and recorded as a video, which is then served to the user.
- We can make the background also move in random directions, to prevent just detecting which pixels are changing and drawing an outline.
- I tried also having the text itself move (bounce like the DVD logo). Somehow makes it even more readable.
I definitely know nothing about how LLMs interpret video, or optics, so please let me know if this is dumb.
yeah - I actually was initially confused since I wasn't having any issues screenshotting it but had forgotten that I have the default site zoom set to ~65%.
Not sure what you mean - I can screenshot it freely that's not the point the point is if you look then at the screenshot you cant discern the text because its a single frame now
This is on MacOS 15.6, Chromium (BrowserOS), captured with the OS' native screenshot utility. Since I was asked about the zoom factor, I now tried simply capturing it at 100% and it was still perfectly readable...
I zoomed out to 90% and could make out something was there but wasn't easy to read. Zooming out further went back to just being noise. I also tried zooming in but with no success. What zoom level did you use and I guess we have to ask the standard what browser/version/OS/etc?? My FFv142 on macOS never took a screen grab like you did
Zooming out before taking screenshot and the text is no longer obfuscated. I tried and confirmed it works. In fact, the text is perhaps even more readable than the original.
It depends how fast or slow your GPU is. I tried it and saw the effect you described, but within a second or two it started moving and was obscured again. Obviously you could automate the problem away.
They even provide the source code for the effect:
https://github.com/brantagames/noise-shader
https://www.youtube.com/watch?v=Bg3RAI8uyVw
The effect is disrupted by introducing rendering artifacts, by watching the video in 144p or in this case by zooming out.
I'd love to know the name of this effect, so I can read more about the fMRI studies that make use of it.
What I've found so far:
Random Dot Kinematogram
Perceptual Organization from Motion (video of Flounder camouflage)
https://www.youtube.com/watch?v=2VO10eDIyiE
- The captcha would be generated like this on a headless browser, and recorded as a video, which is then served to the user.
- We can make the background also move in random directions, to prevent just detecting which pixels are changing and drawing an outline.
- I tried also having the text itself move (bounce like the DVD logo). Somehow makes it even more readable.
I definitely know nothing about how LLMs interpret video, or optics, so please let me know if this is dumb.
This is on MacOS 15.6, Chromium (BrowserOS), captured with the OS' native screenshot utility. Since I was asked about the zoom factor, I now tried simply capturing it at 100% and it was still perfectly readable...
I guess the trick doesn't work on this browser.