CAMERATRACKER FAQ

Match moving is the art of figuring out from an image sequence the position and characteristics of the camera which shot it. It uses the parallax of features tracked within the shot, and requires a sequence shot with a moving camera to do so (but a static shot is easy anyway right?). Additionally, in the case of CameraTracker, it is able to figure out and expose the 3d positions of a number of feature points within the shot in question. The resultant camera and feature points can then be used for 3d, compositing and motion graphics work, to allow seamless integration into the source shot.

Whilst the underlying technology behind these two classes of tools is relatively similar, their abilities and features differ vastly. Planar tracking allows you to identify planes within the image, and get access to the manner in which they translate and deform across the image sequence. CameraTracker (and Match Moving as a whole) allows you to extract from a sequence shot with a moving camera the characteristics and animated position of the camera, as well as the 3d positions of features within the scene. This allows you to create visual effects and motion graphics within the context of the entire scene, as opposed to simply within the plane of a single element.

Rather than attempting to track a single 2d position, CameraTracker attempts to match move your shot (see the first question). It does this by tracking many 2d positions across the entire image and then figuring out from the relative parallax of these moving points the most likely animation position and focal length/film back ratio for the camera with which the scene was shot.

In terms of the specifics of using CameraTracker, we have put together a 14 part, two series set of tutorial videos which can be found on the CameraTracker training page:

CameraTracker Training Resources

Additionally, Tim Dobbert has written an excellent introduction to the fundamental of MatchMoving:

MatchMoving: The Invisible Art of Camera Tracking (Amazon UK)

Check out our training page, which has a 14 part video tutorial series as well as links to a range of third party sites carrying CameraTracker tutorials:

CameraTracker Training Page

And don't forget the detailed manual which is installed when you install the plug-in itself!

CameraTracker does not support this functionality internally, however, since CameraTracker simply creates standard After Effect's Cameras, Nulls and Solids there are a range of free and paid for scripts for After Effects which provide this ability.

These include (and please note this does not represent an endorsement of the tools in question, all third party content is the responsibility of their respective authors):

Maxon Cinema 4d:

• AEtoC4D - For an example of using this tool, check out Marc Leonard's CAMERATRACKER tutorial on AETuts+ as linked to from our CameraTracker 3rd party training resources section

Autodesk Maya:

• AE3D Export
• MoCon

Autodesk 3ds Max:

• AE3D Export

NewTek Lightwave:

• AE3D Export

If you find other options which work for you, please do let our support team know so we can update the list.

This is very dependent on the script in question, however the most common problems include:

• 3D package's world coordinate system differs to After Effect's. After Effects is unusual in that it places it's 3D system centre at the top left of the frame. Other errors may include left to right handed coordinate systems. Correcting for this is again dependent on the script in question, but oft times the parenting Null object the CameraTracker creates can be used to assist. For example, to deal with a centre framed coordinate system, grab the CameraTracker created null, set its position to be 0, 0, 0 and then unparent all created Cameras/Nulls/Solids before doing your export.
• Export script not carrying all information. Many scripts may not bring across every piece of information. Amongst those, commonly ignored pieces of data include comp size, camera film back, and more. To correct for this, simply copy across the offending items.

The objects created by CameraTracker are done so within the relative world space of the parenting (first created) null. As a result, to use this data directly you need to switch the source Null/Solid to reference it's position and rotation data in absolute rather than relative space. Simply unparent the source object from the scene null, and then copy/paste from the source null to your desired target object.

As usual, track, solve, create scene and set ground plane. Create a Null object at this location using the CameraTracker, by selecting the feature and using the menu entry for 'Create Null'. Now, read the above question regarding copying and pasting Position & Rotation data, and unparent the null object you just created. Now create a Text or Shape layer, set it to 3D and either copy/paste over the data from the null, or parent to it. Adjust the scale and orientation as desired and you should be good to go.

CameraTracker has two main modes of operation when tracking. It can either 'render during analysis' or not. The first is meant for quality checking during tracking, and short tracking only. It is susceptible to interruption signals which may be sent by After Effects when it is backgrounded, or when autosave kicks in for example. This will stop the tracking process.

If you intend to background After Effects whilst you're tracking, or you're not interested in watching it as it goes, switch off the 'Render During Analysis' checkbox before pressing the track button. As an added advantage this will track faster than if 'Render During Analysis' was on.

This is either indicative of a poor feature selection, or a poor track. See our 14 part CameraTracker training series for information on how to pull good tracks and pick good features.

CameraTracker Training Page

See our 14 part CameraTracker training series for a walkthrough of using the tool, as well as tips and tricks for common use cases:

CameraTracker Training Page

Such a camera move is known as a nodal pan (strictly, a nodal pan occurs when the rotation happens around the nodal point of the lens, but most normal pans are close enough to be modelled as such). Class 09 of level 1 of our CameraTracker training series deals with handling such camera moves:

CameraTracker Training: Level 01, Class 09

The technique essentially involves switching the Tracking->Track Validation drop down to 'Rotating Camera' and the Solve->Camera Motion drop down to 'Rotation Only' and proceeding as normal.

Switch the Solve->Focal Length Type drop down to either:

• 'Unknown Varying' if you have no idea of the focal lengths involved in the change).
• 'Approximate Varying' if you have a rough idea of the focal lengths involved and subsequently dial them in as an animation to the 'Focal Length' parameter.
• 'Known' if you know exactly the focal lengths and time frame involved and subsequently dial them in as an animation to the 'Focal Length' parameter.

The CameraTracker works better the more information you feed it with, since this can reduce the number of variables in the equations it's solving. The solve twirly contains most of the parameters that can be locked down in such a manner. Figuring out this data requires a little bit of detective work, with sources ranging from metadata stored in the files themselves through to logs from set. Figuring out the metadata depends on the camera in question - R3D for example provide a range of tools with which you can see the metadata recorded, and dSLR sourced mov metadata may be available from the quicktime player's movie inspector panel.

If you know the camera model, look it up on the internet (for dSLR's dpreview.com is a great resource), and transplant the film back sizes over to CameraTracker. Since the focal length calculated is done so as a ratio against the film back, this should normalise the focal length to be in the believable and realistic range. If you have the focal length, switch the drop down to known (or approximate if you have a guesstimate), and then dial in the value to the focal length parameter.