The Moving Picture: HDV Fact and Fiction

One of the acclaimed benefits of shooting HDV for standard-definition output is the ability to shoot a wide-angle view and then pan and zoom within the video without losing quality. I tried this in a recent shoot, a performance by students in my wife’s dance school. The resultant DVD will be distributed to the dance students’ families, and used as a marketing tool for parents considering sending their children to the school.

In terms of setup, I placed the borrowed Sony HVR-Z1U HDV camcorder in the back center, watched but unattended by a friend who was also minding his five-year-old son. Meanwhile I shot B-roll with my Sony VX2000 from front stage left, where I was connected to the theater’s sound system. Overall, while HDV worked completely as advertised, the shoot and subsequent editing revealed three fictions that anyone contemplating this approach should consider. We’ll deal with each in turn.

First is the fiction of the unattended camera. Not to be dramatic, but bad things happen when you leave a camera unattended, or attended by someone who’s inattentive or inexperienced. When some well-meaning parent turned on a glaring spotlight turned on at the last minute, the harsh light transformed the dancing damsels in white dresses (against a flat black curtain) into formless white blobs moving like amoebas on the stage.

Up front, with the VX2000, I rode the exposure controls like a Formula One racecar driver to minimize the glare. For some totally inexplicable reason, it didn’t occur to me that the camera in the back would be experiencing the same problem. While I couldn’t have been in two places at once (the fiction of multitasking), adjusting both cameras’ exposure settings, I could have at least tried the Z1U’s backlight control or similar control. Another problem to try to fix in post.

This leads to the next fiction, the fiction of “fixing” it in post. Shoot something with your camcorder and it’s done; one hour of video takes one hour to shoot. Address it in post and those hours multiply like Retief Goosen putts in the final round of the 2005 U.S. Open. My first stab at zooming in and around the video in Final Cut Pro took four hours, most of which I lost in a crash. Though FCP stabilized when I updated to the latest version of OSX, I didn’t get those hours back.

Disgusted, I turned to Pinnacle Edition, which was painfully slow when zooming in and around the frames, especially compared with FCP on the speedy new 2.7GHz Dual G5. Then I tried to render and discovered that it took 30 minutes to render 1 minute of HDV downsampled to 16:9 DV–about 7 times longer than FCP.

The biggest problem, however, is the tinker factor. To zoom into the video, you use the editor’s pan-and-zoom controls, setting keyframes at each software-generated camera position. The underlying theory of post-zooming quality worked perfectly (especially outside the spotlight), and it was fun making myself look like a camera jockey extraordinaire, with a seemingly supernatural ability to predict the dancer’s next move and maintain perfect Rule of Thirds framing throughout her sequence of grand jetés.

Set the wrong keyframe parameters, however, and thirty frames later, you exclude the two last Village Girls on the right, or the Dancing Flowers in the front of the stage. What will the parents think? Better set a new keyframe or change the last, another five minutes or so of configuring, previewing, and reconfiguring. Overall, the only absolute non-fiction in video editing is that when you can tinker, you will, and the ability to pan and zoom within an HDV frame sets up an irresistible but extraordinarily time-consuming class of tinkering. Great if you’re being paid by the hour, but appalling if paid by the job.

The last fiction is that of DV-comparable HDV support in the various video-editing applications. These will be completely chronicled in our September issue, but here are some highlights. You can’t input HDV and DV on the same timeline in Final Cut Pro, a capability required to edit the input from my multiple cameras. Adobe Premiere Pro produced fuzzy-looking video when downsampling from HDV to DV and lacked multicam capabilities, as does Sony Vegas. Avid Xpress Pro HD has multicam capabilities but doesn’t yet support HDV.

So, it came down to Edition, which, in addition to the long rendering times (like 8 hours per set), somehow mixed up the field order on the HDV source video, producing a flicker artifact that took me 3 hours to diagnose, two hours to fix in all the HDV clips in the final video, and 15 hours to re-render. Make no mistake–HDV is still foreign to all these applications, and you should anticipate significant hours of (additional) tinkering to get everything right.

In short, while the theory of panning and zooming within HDV video turned out to be true in practice, that doesn’t mean it’s actually practical. On my next HDV shoot I’ll find a separate camera jockey to ride the B camera and drive the HDV camcorder myself, judiciously riding the line between shooting the action and getting the big picture. Then, when mixing the two, I’ll pan and zoom sparingly, only when necessary to fix an error or highlight a critical scene.

About Jan Ozer

I help companies train new technical hires in streaming media-related positions; I also help companies optimize their codec selections and encoding stacks, and evaluate new encoders and codecs.

Check Also

Sourcebook 2021 Tutorials Focus Video and Course Production

Every year, the Streaming Media Sourcebook features tutorials to help streaming media professionals perform their …