Company director Ravian Budde works with developer Roger Bolton on the CoreMelt
group of video plug-ins. He talks here about using Lock&Load video stabilization
and rolling shutter reduction software.
Roger Bolton worked for many years in feature film and commercial compositing, and live visuals mixing. In both areas he frequently found a need for plug-ins to address specific repetitive tasks. Over time, he started writing, using and refining his own pieces of software for certain jobs until he had a substantial collection and started sharing them with colleagues. He began distributing a few of these among the visual graphics community and found a good response. Eventually the idea of marketing these as a set of commercial plug-ins developed.
Once he began marketing them, he could develop them further. After noticing the rising interest in DSLR video, based on CMOS sensor chips that often produced a skew or wobble across the video frame as the sensor scans from top and bottom, he decided it would be worthwhile developing a plug-in to address the problem.
Before finding a platform to work from for this plug-in, however, they worked on a video stabilizer with a developer partner in Germany. He had developed an engine to analyse a complete video clip. Instead of dedicating specific tracking points to base stabilization on, this engine efficiently scanned the entire frame. This was released as product under the name ‘Lock&Load’.
At the same time, they continued seeking a means to address the DSLR’s rolling shutter problems and, recognizing a link between the two challenges, decided to address them together. This was how Lock&Load X arose – video stabilization plus rolling shutter reduction.
When tracking video, a user typically defines a point within the frame, which the computer analyses to determine a cluster of pixels in a particular order, and follows the movement of that point through the sequence. It draws a path, rather like a GPS tracker. Based on the tracked movement, it can conclude: “If we want that cluster of pixels to stay in one position in the frame, instead of letting the cluster move along the path, we’ll move the whole clip to compensate for the motion.”
While this is straightforward if you are concerned with one, static cluster throughout the clip but moving shots captured from a vehicle, for example, involve a continuously changing view with no definable tracking point. You would need to define a new point every few frames. Consequently, Lock&Load can base its analysis on the whole scene, across the frame instead of isolated points.
Lock&Load is designed to be fully keyframable because the basis of video stabilization is, first, tracking the scene, then determining the amount of motion in the clip, and moving the whole frame accordingly – that is, rotating and moving the entire clip inside the frame to compensate for that motion, using the frame as a cropping window. The software needs to zoom in on the view to avoid blank edges appearing inside the frame. By zooming in sufficiently, you can keep the frame within the edges of the image. The software moves the clip around relative to the frame, resulting in a perceived stable image.
But what this requires is a certain, averaged level of zoom. Thus a clip with extensive movement, such as handheld footage with a sudden jerk in it, may require a massive zoom. Lock&Load has a tool giving the user control over the zoom at any moment, called Smart Adaptive Zoom, so that you can just keyframe the problematic frames out.
In general, Lock&Load analyses the whole clip, but should you have a scene such as an overhead shot following a vehicle below, the software can help you cope when the vehicle goes out of view under an overpass, for example. The sudden change in pixels may confuse the stabilizer, so you can establish a tracking area for the frames in question and either focus on that area OR on the rest of the frame around it, whichever works better. This tool is very customizable, frame by frame.
Rendering is done on your GPU, which is specialized for graphical data and will give a preview faster than waiting for the CPU. A standard video card is sufficient, such as on a MacBook Pro. The GPU power also means you can set up a background track to run for the clip’s duration while you continue to work on the CPU. Lock&Load currently runs on FCP, Apple Motion and AE, all for the Mac. PC versions are underway.
Lock&Load’s rolling shutter reduction has a special use for motion graphics and effects artists who need odd bits a of video for textures or background. They can use a DSLR to quickly shoot some appropriate HD footage, but this can result in skew during pans from the CMOS chips, especially when capturing vertical and horizontal lines like tile textures or city streets. In these isolated situations, Lock&Load can be used purely for its rolling shutter reduction tools.
The recently released Lock&Load Express, which costs about USD 80 makes a good inexpensive stabilser to augment Final Cut Express for prosumer users who do not need or want to invest in FCP. It does not have all the tools in Lock&Load X, but gives the same performance and is a good match for the tools in Final Cut Express. www.coremelt.com