I use frigate on docker on LXC on PVE. I have a 10-year old Xeon with Intel iGPU that I pass through to frigate as a detector, and it handles object recognition with an inference speed of around 9ms across my 3 cameras, which is much better than when I used 3 CPU cores as detectors. I also pass the same iGPU to jellyfin for transcoding, and it handles both jobs concurrently without too much fuss (but it does get loud!) Worth considering anything 8th gen Intel or newer with iGPU will do fine with a small number of cameras.

I am now! Summit let me subscribe with a single tap, and I prefer the UI to the web version.

Cheers.

Thanks for trying.

Maybe its the username thing - who do we talk to about bugs around here?!

Thanks! That’s… An interesting bug. I’ve added a couple of communities from .ml without issue, so it does seem to be specific to that one community.