Frigate looks pretty interesting. Do you have an estimation of how much CPU it's using for your cameras?

I have about a dozen 4k cameras I'd love to migrate to a much more capable NVR, but doing so with software typically means I go from a 30W NVR to maxing out a 8 core 200W Xeon even without object detection.

Would love to kill my Duaha NVR and toss this on the Proxmox cluster instead.

The Frigate website seems to heavily recommend a Coral card to offload the bulk of the processing from the CPU.

EDIT: looks like CPU still used for video decoding regardless of Coral being available: https://docs.frigate.video/frigate/hardware/#do-hwaccel-args...

I really dig this idea. I think you just gave me my next project.

Could you share which model cameras and NVR you're using? I've been wanting to put something together like this for a while, but all of the solutions I've seen require a Windows box (usually running Blue Iris) for the NVR which I want to avoid

Someone should sell a complete solution to make it easy for non-techies to use. Sell a small computer (perhaps a Raspberry Pi) with the right software installed, the Coral accelerator, and several cameras bundled in a package. Make it clear that this is just a sale of hardware, not cloud data storage. It should be as easy as buying a Ring or a Nest, otherwise it will forever just be tech people's toys.

This is what I was thinking. Is there maybe not a big enough market for non-techies who care about privacy?

It needs to be basically the same as a Ring with just one additional requirement, plugging another box into power and your network to act as the DVR/brains, with a QR code that you scan to link it. Sure that can have more customization options, but the default has to be zero config. Probably gonna need a cloud server for notifying and serving the mobile app without having keys to the video, which would cost money monthly.

There are hundreds of different NVR solutions out there. The consumer ones are... meh. But they mostly work.

The reality is that except for companies, almost no one REALLY cares about it. It's an expense they'd usually rather not have (like backups), so it always gets the short end of the stick.

I tried that product a couple of times on Google ads, and fewer people clicked on "buy now" than my sister's joke product of "lipstick for your pit bull."

People really don't understand this enough to even know that there's a problem. I have explained to a lot of people that "the cloud" just means renting someone else's computer, and they still don't get it.

Google Ads is too aggressive for this sort of product that requires knowledge of many things. The product is niche and would need a simple youtube video series to explain it, with a descriptive title and description text.

Ideally, in the video(s) you can talk about how the viewer can setup the solution themselves, breaking it down into components, why each is important and how it all works.

Also introduce how the viewer can buy the entire solution from you directly, where’ll they’ll get some custom support/troubleshooting or they may just hire you for consulting.

This will function as your funnel to let your target customers find your solution as they search for ways to problems they have self-identified.

If you’d like you can also do retargeting ads on the viewers of your video, as they’d be more likely to buy your solution vs the avg joe.

I'm more hopeful for humanity now that I know my testing methodology was bad. Thanks for the information!

For this kind of thing, you'd want to advertise it on e.g. Tindie - that's where the likely target audience would look for it, I think.

I just wish it hasn’t been almost impossible to get a coral you for the last few years

I was okay with Arlo for several years — primarily because it didn't have a subscription. With Arlo circling the drain now, I think a cloud-less DIY option is an excellent idea.

What NVR? I’m trying to put together my own system. Also, curious - all cameras POE or do you have some that are local wifi instead?