If I implied that ipi was better than PN I did not mean to do so -- I mean the movements are every bit as good, within the limitations of the unit. Which is to say subtle body movements that look like real people.
The advantage over PN is that there is no wearing of any rig, nor any setup process -- you simply press record. But you also have to process the captured video -- it's a two step process, in that you first record the video and then you process it. While the processing is not onerous, once you figure it out (as usual, I was pretty stupid in assuming I knew how it worked, although the instructions weren't all that clear) it certainly is not real time, which could be a factor for many.
As to that latter point, though, I will say that after doing a few days of tests of the various systems I found as a one-man shop that trying to watch what I was doing in a monitor was detrimental to any performance I was giving. I don't actually think the real time aspects of mocap are that significant in that regard (and, let's face it, the best performers in the world, like Andy Serkis, certainly don't watch themselves as they do their mocap). I'm not saying I could never see an advantage in this -- interacting with some environmental objects might be important -- but by and large it is not any kind of factor for me.
Let me also point out that with the iPi you could certainly have body parts pass in front of each other with no issues -- it recognizes arms and legs quite well. With the one Kinect setup it *will* have some issues if you put your arms behind your back, or try turning around, although you can even get away with some of these moves if you are quick or patient enough for the post processing involved to fix the problems.
Finally, I would be remiss if I didn't say that after thinking it over late last night I am leaning towards getting the two Kinect solution. My own position may be rather unique in this regard but let me explain it fully in case it is of any help at all. I think the one sensor version is as good as one sensor can possibly be and, as I said, as long as you live within those limitations (no arms behind the body, no turning around) you will have as excellent a no-finger mocap solution as you can possibly get. It's also extremely cost effective.
I also firmly believe that a year from now the mocap landscape will be significantly different. I can see the handwriting on the wall -- if we don't have under $500 mocap suits and/or Kinect type solutions that do precise capture of BOTH the face and body (iPi hints they are working on face stuff now) I will be amazed. That's one reason I'm not going to plunge into a PN rig at the moment -- I think it's something folks are going to be kicking themselves very soon.
I have a rather significant birthday in January, though, and at my age I don't have much time left in life to wait for prices to drop and things to improve. I want to be able to do things like turns and arms behind the body or at least not worry if I do want them. Plus there's that birthday and a sort of "reward yourself for living this long" mentality. Since I already have one Kinect V2, and since I can get another (plus the cable) for around $150 on Amazon, and finally, since I'd be paying about $100 anyway for the one version, I think I'll pony up the extra $250 and go ahead and get the two sensor version. The examples for two sensor are VERY good -- PN is not better in any way other than finger capture (which I can do with my Leap if I need it -- my tests so far reveal that finger capture isn't nearly as important as I thought it might be). My concern about the subscription model kind of vanishes then because, a year from now, I think the landscape will be very different and if it's not I can just reup (my hunch is, to remain competitive, iPi will have to lower prices. Or I'll get something cheaper and sell my Kinects on eBay :>
The other concern I had about the two sensor version is you need two computers -- not a huge issue for most (certainly not for me -- I have more computers around the house than I know what to do with) and the mechanism to record from both at the same time looks straightforward. But that's what I'll be testing next. If anyone is interested in THAT process I'll be glad to report on it.
(Oh, and both the iPi and Brekel's software produce Human/IK files that import really well into XChange and map perfectly with the Maya model -- sorry if I didn't mention that earlier but that was a big factor to me. So does Brekel's Leap finger capture software but I might post separately about that sometime if anyone's interested).
Alienware Aurora R12, Win 10, i9-119000KF, 3.5GHz CPU, 128GB RAM, RTX 3090 (24GB), Samsung 960 Pro 4TB M-2 SSD, TB+ Disk space
Mike "ex-genius" Kelley