Editing for Slow Motion
In the first edition for this month, I outlined the alternatives for shooting in slow motion. To recount, I'm producing a project at 24fps for web and film output at full-resolution HD, which is 1920x1080@24fps. The project will have regular-motion footage as well as slow motion. I've made the decision to shoot the bulk of the project in 1920x1080@24p, and I've decided to shoot the slow motion footage at 720p60, rather than 1080i60, for reasons outlined in the first edition.
I'll work through two test cases using two editors: Adobe Premiere Pro and Apple Final Cut Pro. The first test case will be the video that I shot with the JVC GY-HM700U in 720p with a recording format of 24p and frame rate of 60. I could also shoot in similar settings using the S&Q mode of the Sony PMW-EX1 and similar camcorders, and in PN mode for Panasonic camcorders.
The second test case will involve video shot at 720p60 with the Canon EOS 7D. Like many camcorders, the 7D doesn't let you set a record format and a capture frame rate, just the latter, so I used 60p.
Again, since I'm producing this slow-motion video at 24p, the sequence preset that I'll use for both Premiere Pro and Final Cut Pro will be 720p/24p. I would edit the slow-motion footage using this preset, then integrate it into my full-resolution HD (1920x1080) 24p project.
Let's start with a quick look at the sequence preset that I used for Premiere Pro. Since the JVC camcorder uses the Sony XDCAM EX codec, I used the 720p/24fps preset for that camera.
Figure 1 shows the Premiere Pro Properties screen for the footage that I shot with the JVC at a recording format of 24p and frame rate of 60fps. Most importantly, because I set the record rate on the JVC to 24p, Premiere Pro shows a frame rate of 23.976fps. I recorded 60fps, and Premiere Pro displays only 23.976 frames of the 60 each second, so the motion slows down. When I drop the clip to the timeline, it's already in slow motionI don't have to make any speed adjustments in Premiere Pro.
Interestingly, though I shot for about 10 seconds (624 frames to be specific), Premiere Pro shows the clip to be 26 seconds long, which is 11 seconds slowed by 24/60, or 40 percent. Though the data rate of the actual captured file is 4.5MBps, Premiere shows it to be 1.8MBps, again 40 percent.
Figure 3 shows the Premiere Pro Properties screen for the clip I shot with the 7D in 60p record mode, which again is about 10 seconds long, or 614 frames. Most importantly, the frame rate is 59.94fps, which means Premiere Pro will display about 60 frames each second. Since I recorded 60fps, and Premiere Pro will display 60 frames each second, this translates to normal-speed playback.
This sounds a bit paradoxical, since 60fps seems like it should be fast-motion, but any time the capture rate and playback rate are the same, it's normal speed. Even if you cranked up a camera to 300fps, if you played it at 300fps, it would be normal speed.
In the context of our 24fps project, this means Premiere Pro will display 60 frames over each second of video, even though the project format is 24fps. Accordingly, when I drag this clip down to the timeline, it will display at regular speed, and it won't look fast or slow.
Rest assured, however, that you can harness those extra frames in the file to produce slow motion. You just have to manually slow the speed down to 40 percent of original using Premiere Pro's speed control.
When you slow down the footage using this technique, be sure to disable Frame Blend for that clip by right-clicking the clip and deselecting Frame Blend. I did some with and without Frame Blend comparisons and didn't really see a difference, but Off is the proper setting.
Now on to Final Cut Pro.
The approach for Final Cut Pro is similar to that for Premiere Pro. I'll edit this slow-motion footage using a 720p/24fps sequence setting, specifically that shown in Figure 6.
When I insert the 24p-record-rate/60fps footage from the JVC into the project, Final Cut Pro reads the frame rate as 23.98fps, knows that it's slow motion, and handles it as such. I don't have to use the Change Speed setting to create the slow motion.
On the other hand, Final Cut sees the footage from the Canon as 59.94fps, and will play it in realtime unless I do slow the speed with the Change Speed control, accessed by right- or CTRL-clicking the clip and choosing Change Speed.
Here, though, I have to be careful to disable Frame Blending, lest I create the faux-interlacing artifact shown on the left in Figure 8. To be clear, if you shot at regular speed (24fps in this case) and wanted to slow the video down, you'd probably want Frame Blending enabled. On the other hand, if you've shot at 60p and have the original frames to useno interpolation requiredmake sure to disable Frame Blending.
For you Clara Peller fans out there (and who isn't, or wasn't?), I know what you're thinking: Where are the video files? Well, calm down, they're coming, after enough caveats to betray my legal training.
First, I decided to test slow motion using two different cameras, rather than the same camera with two different takes at two different settings. The first alternative gives you identical test footage to work with, so the jumps, spins, and jetés all occur simultaneously, with multiple, unavoidable differences due to camera positioning, framing, and different lens/CCD/CMOS/format setups. The second gives you similar but different test footage, with identical framing using the same camera. I chose the former, which enables comparative images like those below, but inevitably means that the two videos will look subtly different.
By the way, the shoot was for an audition clip for Tino Sauter, a rising star in my wife's ballet company whom she'll lose to a full-time school after this year. He's an amazing talent, and we'll all miss him. I shot in my wife's studio using only the overhead fluorescent lighting.
From a slow-motion perspective, the cameras did wonderfully, producing 60fps that slowed as expected in both editors using the techniques detailed above. That said, I wish I had done some things differently with both cameras at the shoot.
With the JVC HM700U, I had the 14X Canon lens option and shot with the iris wide open, gain set to zero, and shutter speed set by the camera, since part of the value-add of the variable-frame-rate technique is that the camera calculates the optimal shutter speed for realistic motion blur. The camera doesn't have autofocus, so I focused before each shot using both the focus-assist mode and zooming in close to the subject. I did not attempt to refocus during the shoot.
Looking at the JVC clip, the focus and exposure look very good (considering the ambient lighting), but the motion blur is evidentperhaps too much so. In a perfect world, I would have experimented in situ with manual shutter speeds to see if I could achieve a setting that delivered a slightly sharper frame.
With the Canon 7D, I used the EFS 10mm-22mm lens, with aperture at 4, which was as low as I could go, and shutter set at 125. I shot in Auto-ISO mode to let the camera determine the best ISO level. I used the Canon's autofocus assist before each shot, and looking at the results, wish I had used the zoom-focus controls as well.
In these images, the focus appears a bit soft, but the motion blur is very limited. The focus softness could have been due to the wide-angle lens, poor focusing on my part, or some combination of both. The video also looks a bit noisy.
In a perfect world, I would have shot with a longer lens with a lower ISO rating, which would have reduced noise and (assuming that it wasn't user error) produced a slightly crisper frame. However, in the relatively shallow studio, I was forced to use the wide-angle lens.
Long story short, what I want you to take from these videos is that both cameras can produce very sharp slow-motion videos if you shoot and edit as described above. You should draw no qualitative conclusions about either camera from these results. Given a different lens, more time, or a better shooter, both cameras probably could have done a lot better.
Finally, to produce the videos that you're seeing, I copied the 720p videos into a widescreen 24p DV sequence preset and adjusted the positioning so that Tino was more or less centered in the frame. Then I rendered out at 854x480 so as not to scale vertically at all. Essentially, this gives you a center cut of the original 720p video. The alternative would have been to scale the complete frame to my delivery target of 854x480, which would have kept Tino in the frame but made him much smaller.