home about gallery publications education links contact

..::bullet time::..

According to John Gaeta, Visual Effects Supervisor on 'The Matrix', Bullet Time is "... slowing down time to such an extent that you really see everything around you as clearly as you possibly could". At the end of the day, it's a method of positioning numerous camera on a rig encompassing the viewed object and all shot at the same time. Ideally two cinematic camera (one at the start and another at the end) should also be used, but depending on the requirements, this isn't all that necessary.

In Keanu's 'rooftop bullet dodge' sequence, 120 cameras and two motion picture cameras were used. For our example, we can start with a few less! It is relatively straightforward to position the cameras to focus on one particular object. Position camera around the object equal distances apart to give a smooth motion when all the frames are composited later, or place them close together at the beginning and ends to speed up and slow down the motion. This can be quite tedious, and the result can appear quite jerky, so unless you've got some equipment to interpolate the frames, the best way would to draw the camera's motion path as a spline.Linking the camera to the spline (and amending the motion if required) it can then be cloned several times by snapshotting the camera so several are produced along the path.

Once this has been produced, you have the tedious task of rendering each individual camera as a single frame (which could be performed easily by automating the task by scripting), to then composite to create the finished animation. If required, cameras should be positioned at the beginning and end of the Bullet Time sequence that match the speed of the Bullet Time motion so the full effect of moving from normal time to Bullet Time and back to normal time is as smooth as possible.

This method is ideal for viewing particle systems due to the fact that even if motion blur is used, the particles (or any other object's motion for that matter) will be blurred only in the direction of the particle's motion and not including the cameras motion, therefore giving the full effect of a snapshot of time. If it's just objects that are concerned, then the multiple camera setup would not be required (unless you want the directional motion blur effect). Simply slow down or freeze all motion in the animation by keyframing manually or adding motion controllers such as a sharp step bezier, circle the desired object, then restart the motion. Again, this can be tedious to set up, especially if a lot of elements in the scene are animated.

To work around this, and also to give the full effect of slowing down and speeding up time, do just that. I'm not on about changing physics laws, but (after saving the initial file), crop the desired time segment (even if it is only a frame or two) by selecting the start and end frames and amending them to the start and end frame values of the segment you have chosen. Then simply scale up the animation to, say, a couple of seconds. Animate the camera around the desired object of focus, making sure it ends up in the correct place for good continuity, render all three segments (start, Bullet Time, end) and composite in Post. Again, particles are not suitable for this method, unless the motion of particles is manually slowed down (on most packages), they will continue at their original speed due to them having manual birth rates and speed presets, and these remain constant throughout.

The Matrix is a good example of the Bullet Time effect, and several people have jumped on the bandwagon, so much infact that it seems to be everywhere nowadays. However, one really good example is Jeff Lew's "Killer Bean" CG short, which can be found at www.flicker-fusion.com.

Enlarge Screenshot

Rendering off each camera as a seperate frame (several overlaid renders displayed here) gives the view of an instance in time. Particles are frozen and can be panned around without disturbing the motion blur.

Enlarge Screenshot

Snapshotting cameras along a spline results in a much smoother motion than if you place them individually. note the tighter camera placement at the beginning and end of the spline: this speeds up and slows down the end motion.

Initially published: 3D World magazine, Issue 6, December 2000.

Copyright Pete Draper, December 2000. Reproduction without permission prohibited.