Multi-Camera 3D Editing

[David Tiede]

I need some help with editing technique using Premiere Pro CS5.

First, here is the setup. I shot 2 hours of continuous footage using 8 Canon Vixia HF21 camcorders simultaneously. The cameras were used in pairs to shoot 3D footage from 4 different perspectives. These cameras are not gen-locked but were started in sync using Lanc Shepherd controllers. The cameras were purposely selected for highly matching frame rates. Two of the 3D rigs ran the entire time within +- 0.001 sec difference in sync between frames (the controller measures this to within 0.01 msec). The other two rigs required a couple restarts to stay within this tolerance.

I put the footage for each pair of cameras in a separate sequence for that pair. If there was any problem with starting and stopping of recording between the pair (very seldom), that was fixed in that sequence. The four sequences from the pairs of cameras were then nested into another sequence. The various clips in the source sequences (since the four cameras rigs were not all started at the same time, for one thing) were all lined up in time with the footage from one of the sequences which ran the entire time. At this point, all four sequences (four for the camera pairs and one which includes those four sequences) have exactly the same timeline. I reduced the scale of the nested sequences to make a 4 camera multi-camera monitor to make sure all four camera rigs were in sync with each other for the entire timeline.

My next task is to perform a multi-camera edit of this footage. This is the area in which I’m seeking some advice and help. If this were 2D footage, I could probably do this with the Adobe multi-camera monitor. However, I’m concerned about the next steps which are a little more complicated. For each cut in the final editing output there will be two camera inputs (left and right). All this footage is in the MTS format provided by the cameras. Eventually I’d like to send each of these cuts (from a pair of cameras) through StereoMovie Maker. This program automatically aligns the two camera views using translation in the x and y direction, rotation about the center, perspective corrections in the horizontal and vertical directions, changes in scale, and setting of the stereo window. After the alignment, the images are cropped to the correct aspect ratio and scaled to fit the output. However, this program only takes AVI or MPEG1/2 files. Thus, after editing, the cuts will have to be converted to something else before input into StereoMovie Maker.

I guess my first question is: what format should that be before input into StereoMovie Maker? I have GoPro Cineform Studio which I’ve used in the past for 3D editing. The Cineform intermediate codec is recognized by StereoMovie Maker. Cineform also has a product called First Light which I’ve used. This product uses metadata to link left and right footage. It also provides tools to do the alignment I mentioned above. However, I find the output of StereoMovie Maker to be far easier to do and superior to what I can create in First Light. So, back to the question, would you recommend something other than Cineform for my purpose?

I’m not very experienced with editing on the timeline. One method I’ve thought about is to find and mark the cut points in the nested sequence which I can see from the multi-camera monitor. How do I cut the appropriate cut from each camera pair onto a timeline? Can I do all this within the nested sequence? Or, should I find the corresponding time in the source sequence and use that to select In and Out points and move them into a Final Cut sequence? I once knew, but can’t remember nor find, a keyboard shortcut which would find the corresponding time on the timeline of the source clip. That is a second question: How do I find the corresponding point on the timeline of a clip in the Source monitor so that I may use that to select the In and Out points?

One thing I can do if the final cuts are taken from the source sequences is to change which of the two cameras is visible in the timeline. Thus, in the end I could make all the cuts show the left camera, render that, then show the right camera, and render that.

Also, I’ll want to do some digital panning and zooming inside the video. The video was shot at 1920x1080 but will be produced at 1280x720. This is equivalent to increasing the zoom by a factor of 1.5. The stereo base used for each camera pair took into account the zoom of the lens, distances to near and far points, and the idea I’d be doing some internal cropping before the final output. I’ll also want to manipulate the stereo window for each cut. This involves shifting the left and right views with respect to each other. Also, some of the shots will include some cameras zooms. These aren’t entirely uniform but I’ve found I can use keyframes to keep the wandering aim point of the lenses under control.

I realize this is a longish post, but hope you aren’t scared off by this and can provide me with some advice.