I switched to the individual #RTSP streams from the ip cams with this #python script rather than use the #Frigate restream function which seemed to be a bit glitchy and also slow. It seems to be working pretty well overall. I haven't seen any failures yet.

I went ahead and switched over the webpage to show photos from this setup. I'll watch it for an hour or two to make sure things appear to be working correctly and then I'll turn off the old computer.

I might add more logging and ...

I've got the script now grabbing stills via #FFmpeg from the #Frigate #RTSP re-stream. All of that is working quite well but it's slow (takes about four minutes to process 16 cameras) and occasionally will fail to open the RTSP stream. There's no pattern to the failures. I tried both on the #Debian terminal container and from my Windows desktop, and it's equally slow and prone to occasionally not getting an image.

I guess I'll try switching it to the direct camera RTSP streams?

It's interesting to look at the log from letting the script run overnight on a 2 min interval. It mostly worked but there were periods of up to an hour where it was not retrieving images and was failing with a 404 error (it's accessing #Frigate via the HTTP API).

I am going to try @ai6yr 's suggestion tonight and instead access the #RTSP streams. I'll start with the Frigate re-streams but I've got a feeling I may end up pulling directly from the cameras...

I know some people might use a video camera with a CAT5/6 cable to import the RTSP video stream into OBS but theses are usually video surveillance cameras that have some limitations.

We can already capture the audio from a microphone in Kdenlive, why not capture RTSP/RTSP or NDI/SRT as well?

https://docs.kdenlive.org/en/project_and_asset_management/capturing_audio.html