This seems even worse to me. Wave at the camera to prove you are human... This is a whole other level of creepy above depressing captchas. I would not do this, ever.
You would need to trust the service the same like people are trusting logless VPN providers with their sensitive internet traffic logs.
Even if the footage was never recorded or deleted instantly after. Even if it was 100% anonymous and 100% guaranteed secure, it still feels like some creepy dystopian nightmare.
A system like this means we've catastrophically failed with technology. It's too dehumanizing. At this point, burn it all to the ground, we're done.
If webcam or mic is not detected and webcam recording or mic recording was given then obviously malicious actor is trying to cheat the captcha and submit fake recording/s.
Edit: You would be prompted with something like: "Wave your hand from left to right." And then video recorder would be presented to you and you would record yourself doing it and review the footage and submit it to the captcha task.
The same applies to mic and voice recording. For example you would be prompted with: Please say: "Life is like riding a bicycle. To keep your balance, you must keep moving." And then voice recorder would be presented to you and you would say it, review it and submit it in order to fulfill captcha task.
I'm betting on a physical recording device that a person has not on the ability or inability of bots or an AI to come up with a human video recording or a voice recording.
A bigger problem is that once you've given some random site permission to access your mic or camera, it can use that access for anything it wants.
And speaking of recordings you can always check metadata and see if it is genuine or fake(edited).
Metadata is trivial to modify, e.g. FFMpeg can do it, and it usually doesn't say anything about genuine-ness anyway. Besides, if you're generating your fake input in software, you'd just generate metadata to go with it.
Sorry, but there are so many holes in your idea that it seems completely unworkable. Even if new developments made it more feasible (new webcams and microphones that cryptographically sign their recordings? Still fakeable by the speakers and screen setup above. Depth-sensing cameras might be trickier.) people would rebel against allowing any and every site to access their webcam and microphone (and hence possibly spy on them).
I imagine it’s awful for people with English as a second language.
Exactly captcha is too much English oriented. I don't have anything against American English or British English but majority of people on the internet are not native English speakers anymore or are not familiar with all the words and variants/nuances of English.
I was thinking how to solve this problem and I came to the conclusion that it is very hard to design puzzle for every country in the world so I decided to put them in cultures something like Civ game but then again the whole process of solving Captcha should be easy not exhausting and time consuming.
My solution was to use webcam if available and make people wave hand or some other gesture in order to verify that you are really human. You can detect webcam presence and webcam input very easily. Or if mic is present you can use mic and pronounce some sentence in order to verify that you are human. As well as webcam you can detect mic presence and mic input easily so Captcha can not be cheated by automated bots.
My solution is humane input as solution to the puzzle or to the task not observing pictures and writing down the text because every computer bot can do it just as good as humans or even better.
Captcha should be easy to do for humans, easy to verify but hard to solve for bots.
In my opinion depressing Captcha is not the problem the problem is that bots came to the point where they are as good as humans solving Captcha but some people struggle solving Captcha because they are not familiar with American culture or simply they do not know English.