Instead I would focus on an issue that you've experienced, perhaps something that you find annoying and can build a solution for. It's highly likely that someone else has the same problem as you and is willing to pay for a solution to it. You get two benefits, one is that it's something that you've experienced so you're probably passionate about building a solution for it, and second it's a problem that exists instead of creating a solution in chase of a problem.
This article gets posted a lot, but I think it's worth a good read.
This blog post helped me a lot as well. It helped me get better as seeing problems and thinking of solutions to them:
maybe combine the badly rated projects with trending ideas from https://insightrend.com/ ?
Also go on reddit, HN commenters don't vent here, but you can find plenty of venting on reddit. You will no doubt find problems to resolve and if that isn't enough. There is even a sub for startup app ideas where people for the most part not capable of developing it themselves come there to "vent" why isn't there an app that does ...... ?
Or... figure out a mechanism that allows power skivving of bevel gears?
Or at least an open source hardware project that allows people to make precision gears at home, so they don't have to pay $50 for something that should be possible for far less.
You can make toy gears with a 3D printer, but for any meaningful amount of torque transmission you need hard contact surfaces against each other which usually means metal, and forged, case-hardened metal.
If your contact surfaces are not perfectly in contact at all times (they never are) then you need something hard and also tough to absorb backlash in the geartrain.
Once you cut the part, then you send it off to hardening, get it back and then grind it to final dimensions, for a really, really good gear.
I want to make it possible for a job shop, or someone in their garage, to be able to turn out any given spur, helical, or bevel gear. I believe the low cost CNC controllers and open source software can be leveraged to make a machine to do this for less than $1000.
Hardening is outside the scope of the project, but grinding is just cutting by different means, and should also be possible.