This seems overly complex. Why require input passthrough?

It seems simpler to make an OBS plugin that way you are able to reuse a lot of work that already exists for game capture and post processing.

I would assume that most of the code is the way it is because "helping users flag accessibility concerns in real-time" in the about implies that they are play testing games using Chroma on top. Using OBS for this would require insane bitrate and tight latency restraints that do not sound very achievable.

Also, at no point does it look like they are actually recording anything. Just screenshots.