I've been thinking about doing something like this forever, but never taken the time to sit down and do it. (And probably couldn't have done a particularly good job of it anyway.) My main usage scenarios are:

- using my hand as a mouse in space-constrained situations where I can't have a physical mouse

- dance class, so the instructor can seek through music, especially for rewinding to the same starting spot over and over again. (This is from the days of in-person classes; it drove me nuts that the instructor had to keep walking back to his phone and then getting it a few beats off)

- spellcasting game

For many uses, I'd be really concerned with false positives. I don't want to have to pin my arms to my sides to avoid triggering something. I'd also want it to be pretty robust to different angles, so I don't have to get my hand exactly parallel to the camera. (Seeing yourself is pretty good for smoothing this over, since at least you'll notice when you're not pointing in quite the right direction. But some applications would be much nicer if they were usable without seeing yourself.)

Overall, this feels more like a tech demo than an application. Sadly, you'll probably have to pick some niche to focus on, because I think the tradeoffs are going to be very different. And dance teachers and yoga instructors don't have much money. (WakaSaba also doesn't exactly suggest the functionality to this English-speaker.)