Find millimeter on Facebook

Related Articles

Shoot Expertise: Multicam Roundup: Part 2

Jul 1, 2006 12:00 PM, By Jan Ozer

Comparing NLE multicam performance and quality.


      Subscribe in NewsGator Online   Subscribe in Bloglines  

Figure 1. Down-sample quality in a ballet sequence.

In the May issue, I compared the multicam features and usability of Adobe Premiere Pro 2.0, Apple Final Cut Pro 5.1, Avid Xpress Pro HD 5.2.4, and the VASST infinitiCam plug-in for Sony Vegas 6.0d. This article focuses on performance and quality of these apps, and summarizes my findings.

Figure 2. Down-sample quality in the resolution chart.

I'll describe the test projects as I go along. First, to test rendering performance, I built a one-minute multicam project in each editor, using four streams of HDV concert footage and changing streams every five seconds. Then, I rendered the project back to HDV format, as you would if you were writing the edited footage back to tape, and then a one-minute DV file, which might be appropriate for DVD authoring.

The test platforms were an HP xw4300 workstation with a 3.4GHz Dual Core Intel Pentium processor and 2GB of RAM, and a 2.5GHz Dual Processor Power PC G5 with 4GB of RAM. The results are shown in Table 1. As you can see, Premiere Pro has a significant advantage in both trials, trailed by Vegas, then Xpress Pro on the Windows platform, and Final Cut on the Mac.

Figure 3. Both videos produced in Premiere Pro from the same footage; on the left in the 720x480 SD preset, on the right in the HDV preset.

But as much as we covet fast performance, most of us prioritize quality over production speed, at least when we can. My next tests, shown in Figure 1 and Figure 2, isolate the quality of two videos downsampled from HDV to DV.

I captured the videos in each editor and set up another project, panning and zooming within the HDV frames and outputting in DV format. Just to be clear, this involves scaling the HDV footage (1440×1080i pixel resolution) down to 720×480 resolution, with both formats interlaced.

Figure 1 shows good quality all around, indicating Premiere's speed advantage doesn't penalize quality. Let's see if we can verify this in the resolution chart trials shown in Figure 2.

Table 1. Length of time (min:sec) to render a one-minute multicam test file comprised of four HDV streams to the designated formats.
For a larger image, click here

Again, Premiere Pro does well, with horizontal resolution (shown via the vertical lines atop each frame) almost matching Final Cut and Vegas, and ahead of Xpress Pro. It appears that Final Cut and Vegas both produce better vertical resolution (shown via the horizontal lines on the left of each chart) than Premiere Pro, with Xpress Pro again at the end of the pack.

That said, these differences are very modest, especially compared to others you'll see later in the review, and they probably wouldn't be noticed by most viewers, even with the benefit of side-by-side comparisons. If you're downsampling from HDV to DV, all of these editors do more than an adequate job.

Our research revealed one critical tip when working with Premiere Pro: Output quality depends upon the project preset chosen at project inception. Specifically, we produced much better results outputting at SD resolution when using the SD preset than with the HDV preset, which is shown in Figure 3.

As with all the tests, we produced both files from the same HDV source footage. On the left of Figure 3 is the video produced in a project using the 720×480 SD preset, and on the right is the HDV preset. The SD preset produced much higher results, which is anomalous, but verified in multiple tests. Clearly, when producing downsampling with Premiere Pro, it's absolutely critical to use a project preset that conforms to your target output resolution and aspect ratio.

Figure 4. DV video upsampled to HDV video.

Upsampling tests

If you shoot with HDV, sooner or later you'll have to upsample DV, or other lower resolution files to HDV format, simply to mix old and new footage in the same project. Unlike downsampling, where editors throw pixels away, during upsampling, editors invent pixels, which makes the task more challenging. Not surprisingly, we saw significant differences in the output quality produced by the different editors.

In these tests, we used two 4:3 resolution DV videos: one is an interview of congressman Dick Boucher and the other is of the same resolution chart shown in Figure 2, but it's shot in DV rather than HDV. The total length of the video file was 110 seconds (1:50) and processing time for each editor is shown in Table 2.

Premiere Pro was the clear performance leader, though, as shown in Figure 4, in this operation, it came at a significant cost in quality. In these tests, Final Cut Pro is the clear winner, with Vegas and Xpress Pro HD roughly tied for second place and Premiere Pro bringing up the rear.

Figure 5. Upsampled quality with Premiere Pro, After Effects, and After Effects using the AlgoSuite scaling filter from Algolith.

Unlike our downsampling tests, the difference is so significant that it's tough to recommend using Premiere Pro for this operation; viewers would notice. Fortunately, Adobe's new Dynamic Link feature makes it simple to use After Effects for this operation by allowing you to insert an After Effects project into Premiere Pro. The results are shown in Figure 5.

On the extreme left of Figure 5 is the image produced in Premiere Pro. The middle image shows the same video scaled from DV to HDV using After Effects rendering engine via Dynamic Link, which produced a noticeably better result roughly matching the quality produced by Sony and Avid, but it increased rendering time to 11:58 (min:sec).

On the far right is the video produced by AlgoLith's AlgoSuite scaling plug-in, which we accessed as a plug-in for After Effects, but it is also available for other tools, including Apple's Shake. While the After Effects image still trails video produced by Final Cut in quality, AlgoSuite delivers relative parity.

Table 2: Times to render a 1:52 DV file to HDV format.
For a larger image, click here

Table 3 summarizes the findings from this article and from part 1, which you can read online at digitalcontentproducer.com/desktoppost/ depth/multicam_roundup_part.

In terms of pure multicam, Premiere Pro should meet the needs of most small producers, though the inability to pan and zoom within HDV clips is frustrating. The more I use the tool, the more I like the way Premiere synchronizes clips and easily flows through color correction and other adjustments made to the source clips into the multicam clips. Performance is top notch, and downsampled quality is very good, so long as you use the right project preset. Sub-par native upsampled quality is offset by great integration with After Effects, with absolute top quality available via third-party plug-ins, an advantage that not all competitors can claim.

Table 3: Multicam and scaling summary. *1=fastest performance
For a larger image, click here

Apple's multicam features offer a unique combination of usability, functionality, quality, and reasonable performance, especially if you only shoot in one format. If you use multiple formats in your project, you'll have to convert to a single common resolution/codec, but at least Final Cut's scaling quality up and down is excellent. I adore the creative possibilities enabled by unlimited pan and zoom and the ability to edit the multicam footage freely after selecting the cameras.

Avid's multicam tool is functional and mature, but challenging to learn, and it's the most expensive editor in the group (Of course, if you're planning to work your way up to Avid's higher end tools, Xpress Pro is also an investment). While Xpress Pro's multicamera color-correction tool is a dream, it's disappointing that you can't easily create overlays and PIPs from the multicam footage. Although Xpress Pro's upsampled video quality didn't measure up to the quality of Final Cut Pro, it was better than Premiere Pro and comparable to Sony Vegas.

As the only third-party plug-in in the group, the VASST multicam tool has somewhat of a Rube Goldberg feel, but it's inexpensive and gets the job done, with no significant negatives. Vegas' performance, once a significant competitive disadvantage, is now a neutral to positive, especially if you can set up multiple rendering nodes on a network. At a combined price of about $520 for Vegas and the infinitiCam, this is the most inexpensive alternative for producing mixed format multicamera editing.


To comment on this article, email the Digital Content Producer staff at dcpfeedback@prismb2b.com.

Share this article




Continue the discussion on “Crosstalk” the Millimeter Forum.


© 2012 NewBay Media, LLC.

Browse Back Issues
Back to Top