Grimfoss | 🔜 TwitchCon

I want markerless mocap to succeed as much as anyone, but I’m not backing Soulart yet. Why: 1. Single-sensor physics. I’ve worked with CV body tracking, stereo, and ToF. Metric depth helps a lot, but one viewpoint still cannot see a limb behind your torso, a turned back, or crossed arms. Kinect proved a single depth sensor can do skeletal tracking in 2010, and it also proved exactly where that breaks. Nothing shown so far explains why this is different. The demos look curated. 2. Price vs. the alternatives. At the per-unit price ($450), a full SlimeVR set works fully occluded, and a two-station setup (the minimum that survives occlusion) costs about what 16 Slimes plus finger-tracking gloves cost. CV finger tracking is hard even up close in good conditions, let alone across a room in real time. If single-station tracking has the usual optical quirks, the jump over a free webcam pipeline may not justify that. 3. The hype is incentivized. Their own blog says the ~20 early-access testers were required to post avatar footage online, and purchasers get a free upgrade to four final stations. Not a scandal, but I weigh the glowing posts accordingly. A lot of the QRTs and replies read the same way to me, though I could be wrong on that. Nitpick: if pose estimation really runs entirely on the device, a single station should stream VMC straight over USB with no app in the loop. SteamVR drivers and multi-camera fusion need software, fine, but the basic path shouldn’t. Slimes and UDCAP gloves need a middleman because the hardware can’t solve the pose itself. Soulart’s can, so the app should be optional. As a streamer, the last thing I want when I go live is yet ANOTHER app I have to leave open. I’d love to be wrong. If it ships with the latency and precision they claim, it’s a game changer. Until independent raw tests drop, I’m watching from the sidelines.

To chase away the crippling social insecurity, I will be streaming MORE this week! Holy shit is that a schedule?