Face v1 FAQ

• Kinect for Xbox 360 or Kinect for Windows v1 sensor, including its power supply
• Windows 10 / 11 (64-bit)
• Intel Core i5 or equivalent CPU
• 4 GB RAM
• Dedicated USB 2.0 bus for the sensor
• Graphics card with OpenGL support

Both the Kinect for Xbox 360 and the Kinect for Windows v1 are fully supported. The Asus Xtion and PrimeSense Carmine sensors are not. Specs, and how this sensor compares to newer hardware: depth sensor comparison.

They track equally well. The ‘for Windows’ version adds near mode, which can track between 40 cm and 3 meters instead of the regular 80 cm to 4 meters (mainly interesting for face tracking, since closer means more detail), and it adds control over the video camera’s exposure and white balance.

The sensor dates from 2010 and has long been out of production, so it is second-hand only. It sold in enormous numbers though, so they turn up regularly on eBay and similar marketplaces, usually for very little. Make sure the power supply is included: the sensor needs it and cannot run on USB alone.

The official Microsoft Kinect drivers/SDK. The installer automatically downloads and installs these for you, or you can grab the Kinect SDK 1.8 yourself.

It is not a high-end facial capture and retargeting tool; other products exist for that at higher price points. What it does provide is robust tracking that does not need to be pre-trained to a specific actor, at an affordable price.

A single instance of the app uses one sensor at a time. You can run the software more than once and choose which sensor each instance uses if you have several connected.

• Head position as x, y, z coordinates
• Head rotation as x, y, z euler values
• The face points of the template mesh (see the 3D window) as x, y, z positions (not all of these contain unique tracking data)
• Animation units: animated values representing facial poses, for example to drive blend/morph shapes
• Shape units: static values representing facial structure (head height, eye separation, mouth width, and so on)

Eleven, which you can use directly to drive blend/morph shapes in your face rig: Brows Inner Up, Brows Inner Down, Brows Outer Up, Brows Outer Down, Lip Stretch, Lip Kiss, Lip Corners Up, Lip Corners Down, Upper Lip Up, Upper Lip Down and Jaw Open.

Shape units are static values describing the shape of the face, for example how far apart the eyes are. Animation units are animated values describing the expression of the face, which can drive blend/morph shapes in your face rig.

Correct, not all points are tracked individually; internally they are derived from the animation and shape units.

No. These are a direct result of the Microsoft face tracker in the Kinect SDK and cannot be altered.

A single Kinect v1 does not provide enough data to do both well. The limited resolution gives poor face tracking results at a typical body tracking distance, and with a single viewpoint you would have to keep facing the sensor, which severely limits the motion you could perform. Brekel Body v3 does combine body and face tracking, using modern sensors.

• Make sure the video image is well lit and not too dark
• Adjust and play with the Filtering sliders
• Get as close to the sensor as possible (a Kinect for Windows sensor with near mode enabled lets you get even closer)

Video output depends on which video codecs are installed on your machine. When the ‘Ask for codec’ option is disabled the XVID codec is used automatically if available. If you are having codec trouble, the K-Lite codec pack and/or Xvid are recommended; install the x64 version of the codec since the app is 64-bit.

I am no expert, but I believe it is. There is a lengthy discussion about it on laserpointerforums.com.