it definitely has custom region based motion detection, I can't remember specifically if you can create multiple regions on one feed but I would think you can. If not, I'm sure the dev would be receptive to it (he's implemented everything else people have asked for so far, it's quite new software)
Side anecdote: I used MythTV as a DVR/media PC app for probably over a decade. It always felt "fragile" to me, and once I got it working, I was afraid to touch for fear of breaking it again. Someone on this forum suggested Plex Media Center as a replacement. At first I was like, "nah, the devil you know is better than the one you don't." But curiosity got the best of me, and let me just say that I'm now a lifetime Plex Pass subscriber!
Unfortunately, the experience with ZoneMinder versus Shinobi was not the same. I gave Shinobi a fair shake, but in the end, it just felt way too "beta" for me. It's funny how people seem to either love or hate any given video surveillance platform. Before I ever even tried ZM, I was pretty scared, as there are a lot
of negative stories about it on the net... but my experience has been remarkably good. The software has just worked as I expected, and all the while been completely stable.
My only complaint with ZM is the very real CPU- and IO-heavy aspect of its architecture. My cameras output h.264 native streams, and yet ZM converts this to a series of JPEGs. While I'm not hurting for compute resources, I'm always game for efficiency (and a power bill reduction). That is what convinced me to give Shinobi a try.
Sadly, it just didn't cut it for me. I found numerous cases where the documentation was incomplete or woefully outdated. For example, using sqlite instead of mysql, the different stream formats in the web gui, motion detection. It appears to me motion detection is now builtin to the software, but can also be done through one or more plugins. There's no discussion of the pros/cons of the various methods or to help one decide how to choose. The motion detection doc is terse at best, has no screenshots or examples... it could
be as powerful as ZM's, but without screenshots or more detailed docs, there's no way to know short of trying it all. So much for a "quick test".
The web interface for me was terribly flaky. Feeds would randomly go black for no apparent reason. Different stream types would be more reliable, but some would cause a dramatic CPU load that was even worse than ZM. This doesn't inspire confidence.
I have quite a bit of cameras (axis cameras putting out h264), and since I have shinobi set to "stream copy" storage method, it just straight dumps the native h264 streams to disk. no encode/transcode, so vm load is essentially zero (just some disk writes)
So it sounds like you're not using motion detection? That's the killer feature for me for ZM. These days I only periodically spot-check my cameras (as I'm fortunate to live in a neighborhood where cameras really aren't necessary), but when I do---or if there ever was an actual incident---it's nice to have the footage "pre filtered" by the motion detection algorithm. It leaves me with footage of only potentially interesting events, rather than countless hours of complete inactivity.