Voice commands and Security Spy
Gonna get a mini for my NVR upgrade and hoping it can have a 6-8 year life. Voice commands, I'm guessing is going to be one of the next big improvements for NVRs.
1) So is this coming in 1-2 years or in 6 years?
2) Should I get a powerful Mac PC or a basic one and upgrade when voice commands mature.
Comments
-
one voice command could be.
"hey what did my cameras see last night"
Possible responses
1) nothing eventfull
2) A tree fell on the fence
3) Police were at a neighbors
4) during last night rain, the left back downspouts was slow releasing water
5) someone was walking back and forth in front of the house
6) Compressor sounded abnormal
7) ect
-
Ok, sorry last post on this. Will Grok Bot running on same MacMini be able to do the voice command things I was curious about?
-
Just to follow up on this, I like the idea in general, although it's not quite achievable as described - the kind of searches you are proposing necessitates a very high-level understanding of events in the video footage, which is well beyond the capabilities of any AI model that you can run on a regular Mac. The alternative, which is to send video to a cloud AI provider running a much larger model, would give better annotations, but cost and latency would be major issues of doing it that way, not to mention that this goes against one of the major selling points of SecuritySpy that users appreciate, which is that all their data stays local.
That being said, we are actively working on a feature that is similar to what you describe, which is "semantic search" of your recording history. The kind of search you can do under this scheme will be something like "person wearing dark shorts" or "yellow truck on driveway". This will involve an LLM that runs on your Mac and annotates all captured footage, an efficient database that provides fast retrieval, and natural language processing to interpret the request correctly to provide good matches. This feature is coming soon to SecuritySpy.
Voice control could then be a layer on top of this, but the bulk of the work is the annotation/database/language system, not the voice control per se.
If you are interested in upcoming AI features in SecuritySpy, the best thing you can do with your Mac purchasing decisions is to get as much RAM as practical. These kinds of AI models are very RAM-hungry.
-
thanks for reply and update, yeah just asking for a general status (not getting frequent texts) to learn what is going on around the house may be useful and with the help of AI may achievable at some level. though words are easy for me to say.
Haven't update yet (from BlueIris) Been learning HA and realized that Zigbee is worth a look for simple stuff cause thead has been difficult but I expect its my fault, and it all has been consuming my time. Also was waiting to see apple's new stuff, which as expected isn't 24/7 recording and that means lots of missing data, cause as we all know triggered recordings miss useful data cause it doesn't trigger 24/7.
-
It's true that motion detection can miss events - while SecuritySpy's detection is highly accurate, there's always an inevitable tradeoff between false positives (incorrect detections) and false negatives (missed events). If it's important to you that absolutely no events are missed, then 24/7 continuous recording is the way to go, although that uses much more storage (see our Systems Requirements Calculator to get estimates for storage requirements).
In terms of Home Assistant and protocols: Zigbee offers a wider range of accessories, so that's the one that users tend to default to. If you want Matter too, it's possible to run this in parallel the right hardware, but for simplicity and robustness of your mesh it's best to unify on one of these (in my own home, I run a Home Assistant Green with a ZBT-2 that does Zigbee, plus various WiFi devices added via built-in integrations, like HomeKit and Shelly).
