Why would you disagree on the other options? I quickly built a proof of concept Capture device with webcam-capture as the underlaying technology. It works pretty well on my Windows and Ubuntu (MacOS is currently a bit more difficult as you know too), and is only 1.8 MB in size.
https://github.com/cansik/webcam-capture-processing
That’s enough for most of my use-cases and I think for most of the Processing users, a simple library for video capture would be sufficient.