Using AutoTrack to track a video

AutoTrack is a feature in Analyzer that allows for completely automated tracking of the face.  Our more than a decade of experience in facial motion capture has allowed us to learn from tens of thousands of different faces and incorporate them into the different tracking models.  With the universal human face models, we can track the movement of any face, in almost any camera or lighting condition, completely automatically. 

Additional Notes: 

• AutoTrack is a powerful feature in Analyzer, but will never quite get the levels of accuracy of tracking manually with training frames. We recommend AutoTrack for use with pipelines that need tracking incredibly quickly and for characters that do not necessarily have the focus in shots, like background characters.
• Whenever AutoTrack is used, it will automatically overwrite any tracking results that have been created for the entire shot, both from any previous AutoTrack attempts or from manual tracking. If there is a face group that you wish to preserve, select all of the landmarks in that group and hit 'CTRL + L' to lock them. Locked landmarks cannot be overwritten in any way until they are unlocked (select them and hit 'CTRL + L' again to unlock them).
• AutoTrack is unsupported if you are working with Jaw and Cheeks groups. If you wish to track these areas of the face, you will need to track the face manually. 

In this article:

Using AutoTrack

Under the 'Track' file menu you'll find an option for 'AutoTrack' that will bring up the following window:

1: Select a Tracker

Under 'Select Tracker', from the drop-down menu select your desired tracking model from three pre-built universal trackers:

• Static Cam - The StaticCam tracker is designed for use with stationary cameras (see example image below). The actor's face can take up a varying amount of the total frame and can move around. Due to its flexibility as a tracker, this is the default tracker selected and can be used with virtually any facial footage. This includes normal static camera shots or even shots with multiple faces, though you will likely need to use a Neutral Frame for more complex shots (see Step 2:Neutral Frame below)

• Headcam Color - The Headcam Color tracker is designed to be used specifically with color videos from a head mounted camera. To work properly, the video must be in a vertical orientation (more tall than wide) and the face should be occupying the vast majority of the frame like in the example below. Due to the stringent requirements for this tracker, if you are getting errors or an inconsistent track with either Headcam tracker, try using the Static Cam instead to improve your chances of getting a good Autotrack result.

• Headcam Grayscale - This tracker is almost identical to the Headcam Color tracker except for being specifically made for greyscale (black and white) videos and should only be used for that type of footage.

We recommend running a test with your video with each of the available trackers to determine which gives you the best results for your particular video setup. If you are unsure of which one to use, select the Static Cam tracker as it is the most versatile of the three.

Step 2: Neutral Frame

While not required, supplying an accurate Neutral Frame (.fwlf) from Analyzer will in most cases significantly improve the accuracy of AutoTrack. You can find detailed instructions for creating and using Neutral Frames in the following article: Neutral Frames

Step 3: AutoTrack

After selecting a tracker and optionally providing a Neutral Frame, simply click 'AutoTrack' to begin tracking.  The process is quite fast and should only take a few minutes at most.  When the progress bar completes, you'll have tracked results on your timeline in Analyzer.

Improving your AutoTrack Result

To improve the tracking you get with AutoTrack, you can use the automatic results to quickly setup training frames and generate a tracking model for your actor.  You can also move landmarks by hand, or delete any "messy" sections.  Another option is to utilize the AutoTrack results for a particular face group (Mouth, Eyes, or Brows), Lock the points that are accurate, and only edit the groups you're unhappy with.

In most cases, the tracking should be accurate enough to bring directly into animation without needing editing or cleanup by hand.