Body v2 FAQ

• Kinect for Windows v2 sensor, or Kinect for Xbox One with the separately available USB 3.0 / power adapter
• Windows 8 / 8.1 / 10 / 11 (the USB stack of Windows 7 and older cannot handle the bandwidth of the v2 sensor and is not supported)
• USB 3.0 port, Intel and Renesas chipsets only (other brands may or may not work)
• DirectX 11 capable GPU (even old ones like an Intel HD4000 or GeForce 610M work)
• 4 GB or more RAM
• Dual core i5/i7 CPU around 3 GHz or faster (slower works but may drop frames)
• 1280x1024 screen (1920x1080 or higher recommended)

The sensor dates from 2013 and Microsoft ended Kinect production in 2017, so new units are no longer sold. Second-hand sensors are usually easy to find on eBay and similar marketplaces. The Kinect for Xbox One and Kinect for Windows v2 are functionally identical; for PC use the Xbox One version needs the separately sold “Kinect Adapter for Windows” (also second-hand only these days). Specs, and how it compares to sensors still sold new: depth sensor comparison.

Yes and no. Tracking starts with joint positions from the SDK, but a lot of technology has been built on top of this to improve quality, especially for rotations and additional joints.

No. Those first-generation sensors are covered by Brekel Body v1. Brekel Body v3 supports both Kinect generations, plus newer sensors.

No, Body v2 is built around the Kinect v2 sensor and its SDK. Body v3 supports a range of other sensors, including RealSense and plain webcams.

Yes, Body v2 happily co-exists with the other Brekel apps, and the Kinect v2 sensor uses completely separate drivers from the v1 generation, so nothing interferes.

No. The three sensor technologies use slightly different infrared light frequencies for their depth measurements and can be used together.

These are the chipsets officially supported by the Microsoft drivers; they are proven to handle the high bandwidth requirements robustly. Other brands may work, or may be unstable.

The Microsoft drivers convert the raw infrared stream on the GPU to ensure real-time performance, using DirectX 11 DirectCompute code.

No. Besides the bandwidth requirements, there is a Microsoft driver/SDK limitation that prevents access to more than one Kinect v2 sensor per machine, even with extra USB 3.0 cards installed. There is no way around this.

Partially. All Brekel apps have a Record Triggering feature that syncs recording between multiple apps on one or more machines, so you can record from several sensors simultaneously. Body v2 does not calibrate or fuse that data into a single skeleton though. Body v3 does exactly that: aligning and fusing multiple sensors, of mixed brands, across machines.

The v2 sensor measures depth on a different principle with about 3 times the fidelity. Skeleton tracking uses 25 joints (vs 20 on v1) with more anatomically correct placement, joint rotations are calculated more accurately, and data is generally less noisy. Up to 6 people can be tracked simultaneously, and the wider lenses see a bigger area.

A person can be tracked between roughly 0.5 and 4.5 meters from the sensor.

The software requires Windows. It has been confirmed to work under Boot Camp on Intel Macs. Virtual machines usually do not work since they lack DirectX 11 and full USB 3.0 support, and Apple Silicon Macs have no Boot Camp.

• Make sure your graphics drivers are updated to the latest version from the NVIDIA/AMD/Intel site
• Place the sensor at about chest/head height, slightly tilted down so it can still see the floor
• Hand and forearm rotations depend on visibility of the thumb
• Acting with open, relaxed hands works better than fists
• Avoid baggy clothing that obscures the wrists and hands
• Some clothing reflects very little infrared light, which lowers tracking accuracy; check the IR/depth views
• Tune the Forearm and Hand Roll Sensitivity settings in the Skeleton tab to trade stability against fidelity

Yes, see updates and upgrades in the shop.

Body v3 is the current version and in active development: it supports many sensor types (including Kinect v2), can align and fuse multiple sensors, and adds an optional deep-learning tracker. Body v2 is the cheaper option if a single Kinect v2 sensor is all you need. See the Body family overview for a comparison.