I need picture in picture in cameras stream. Also it will be nice to have join five video stream to one stream.
HA is really only a command and control server or monitor and control server.
There is no real heavy processing going on in it. This is put out to other software to handle.
The required processing power to decode 2 video streams, scale one of them to embed it and then encode it again require a lot of processing power, especially if you want it to be just somewhat real time.
You should probably find some separate software for this and have a good hardware platform for it.
I do not know if Frigate can do it.