I store in home Depot containers with desiccant in the bottom. And have another modified to print from. (Printed spool holder with bearings, reverse Bowden through container side)

It really didn’t change my workflow, just where to get the spools to load, and where to load them.

Print quality improved greatly.

Handle is much different than efficiently calculate.

Just make sure they have substreams and get a coral (or GPU) for detection. Then a pi could probably handle it.

I have enough processing to do it all on CPU (8 cams+doorbell) but it ramps up the power usage, so it was better to use the gpu I already had for transcoding, as detection.

Without you can still just record and overwrite. Not that it’s extremely useful without detection and notifications.

Yes, the issue I have with no sub stream (only on the doorbell) is that it uses more processing for detection on such a large resolution.

The doorbell does, except for no sub stream. And the only way for mine to be setup is their bs app.

I should’ve looked better when buying, but alas. I have this one and I’m lazy.

I have a few. Some try and call home (mostly the doorbell, every 10s). The others are easy to setup and run with frigate.

There are screw together butt connectors, in my experience have a more solid connection than the crimp style as far as pullout is concerned. https://www.posi-products.com/index.html

Just… how?!?

Best budget printer is one someone is selling.

Buy it, then if you have to, upgrade it into what you need.

My solution to the same problem was to disassemble the echo. Take the speaker output and wire it to a cheap amplifier (adafruit pam8302), and then wire that to a larger speaker in a 3d printed enclosure.

Only thing I wish I did was wire the shutdown pin to one of the extra pins on the echo to turn off the amplifier when nothing is being sent to the speaker.