Would it be feasible to do picture-in-picture in hardware? Digital video feeds, not super high-res, but needs to be updated preferably 120 times a second. Are there any single-chip solutions to this? I know text is pretty easy, but video sounds more than a bit trickier.
If you can wait til march, then one of these should do: https://www.crowdsupply.com/numato-lab/opsis
If you can't, then a bog-standard PC with capture cards would work, though latency would be a challenge. Or not, depending on your requirements.
you're looking for a thing called a multiviewer.
they're about a hundred fiddy bux for a basic unit that takes 2 1080p hdmi in and gives 1 hdmi out with pip and vertical split modes.
120hz frame rate is taking the piss though, there isn't even a standardised format for that.
Whats your source? Whats your production workflow? Are you even producing? Whats the target output?
If you're bringing two streams into a computer then open the streams with ffmpeg, use a PIP filtergraph, and pipe the output to a virtual video device to be captured elsewhere