Author's note: this article was published in December 2003 and all programs have released new updates since that time. Links to more current reviews are at the bottom of the article, but this seminal piece sets up how we compare professional codecs in these very important skills.

With landmark recent releases from Adobe, Apple, Avid, Pinnacle, and Sony Pictures, today's video-editing scene boasts a surfeit of showstoppers. But which one is truly today's premier prosumer NLE? It all comes down to performance, and here we stop the show, break it all down—from keyframes to filters, ease of use to output—and determine who ranks first among the current fab five.

By Jan Ozer - Originally published in EventDV, www.eventdv.net

Video editors are like elephants; your impression is largely formed by how and where you touch the beast.

Even worse, to a large degree, your impressions are formed by what you use and what you know. Even the most objective reviewer would have a hard time characterizing an editor he or she has worked with for many years as unintuitive, while new editing environments always seem foreign and forbidding. When focusing on objective comparative criteria, like output quality, how do you account for the reality that a reviewer will generally prove more proficient with a familiar program, and know which knobs and dials to turn to produce absolute top quality?

In our search for an "objective" comparison of video editors, here's what we came up with:

• We selected five of the top prosumer editors: Adobe Premiere Pro (MSRP $699), Apple Final Cut Pro 4 ($999), Avid Xpress Pro ($1,695), Pinnacle Liquid Edition 5.5 ($699), and Sony's Vegas 4.0d ($559).

• We created four multipart projects in each program, focusing on critical functions like video overlay, color correction, slow motion, and image stabilization.

• We sent the project files, assets, and our results files to each company with a detailed listing of problem areas that our tests had uncovered, inviting each company to call or write with their comments.

• Four of the five companies responded with corrections, explanations, and admissions. The only company who didn't respond was Apple, though provided the same lead time as the other companies.

Then, we compared the output quality of the files produced. This is what we found.

Overlay Tests
Video overlay is a capability that allows video editors to place portions of one video over another, or to impose logos and other images over videos. The most common technique is to shoot one video against a blue or green screen, and then combine the two videos, using a technique called "keying" to remove the green or blue background. This is how the weatherperson appears over the weather map on your nightly news.

All tested programs support video overlay, though feature depth varies to a degree. First, most offer a number of keying techniques, including chroma or color keying, which removes the background color, luminance key, which keys on differences in the brightness of the clip, and alpha-channel keying, which removes the alpha channel from the video or still image. Our tests analyzed color keying and alpha-channel keying.

With color keying, there are two basic levels of features. Typically, you use the eyedropper to choose the background color in the clip to be overlaid that tells the video editor which color to key out or eliminate when combining the two clips. Most programs also have a color tolerance value that increases the range of colors eliminated during keying, and edge controls that soften or feather the edge of the overlaid clip.

(See Figure 1 to see basic color keying functions we used in our tests.)

The second layer of features consists primarily of spill suppression, which converts the background color to grayscale, which blends more easily than the distinct blue or greens used for most real-world keying. Basically, the theory is if you can't get rid of all traces of the blue or green background, make it as unobtrusive as possible. We used this feature when necessary and available.

Other advanced features include secondary color keying, which allows you to replace one solid color in the video with another; for example, converting a yellow car to red. We did not test secondary color correction, but note its availability in the Overlay features table.

Our overlay tests consisted of three ten-second segments, each with three layers, with three videos requiring chroma keying, two spinning logo files with an alpha channel, and one spinning EMedia logo. Two of the three chroma key files were DV source footage, with the other a 270x240-resolution raw AVI file zoomed to 2X before keying. We created the spinning logos in Ulead's Cool 3D Studio and downloaded the EMedia logo from www.emedialive.com.

TABLE 1: Chroma key features, performance, and rating

Vegas was our overall winner in the Overlay category. Vegas' edges were almost completely alias-free, even during normal speed playback, with only negligible degradation on the logo file. Vegas does have a tendency to darken the background video, which we watched for and corrected, and the edges of the video were slightly messy, which we cropped out with the 3D PluginPac freely downloadable from www.debugmode.com. Logos were crystal clear throughout, and there was only a hint of aliasing on the zoomed final shot. The only negative was Vegas' well-known slow rendering times, though definitely worth the wait here.

Edition was very close to Vegas in quality, removing all of the blue in most shots, with good detail and only slightly more jaggies than Vegas, which is most evident in the dancing girl clip, as is a noticeable degradation of the spinning logo. Color integrity was very good with a great range of tools and great speed due to both real-time filters and background rendering.

Avid Xpress was next, doing a good job eliminating most of the key color, with spill suppression necessary on all three sequences. Edges were generally smooth with virtually no degradation on the spinning logo. The only negative was a tendency to fade the clips slightly that we corrected on most test clips by boosting both brightness and contrast.

(Click Figure 2 to see a full resolution sample comparison of smoothness in our overlay testing results.)

Results with Final Cut Pro were suboptimal. We never completely eliminated the blue from the test clips, despite trying all available keying methods. Though spill suppression converted the blue to gray, which was less obtrusive, jagged edges were evident in the dancing girl clip. Smoothness controls tended to eat into the overlaid edges, limiting utility, and producing noticeable jaggies in the car scene, especially during live playback. On the plus side, Final Cut Pro did little damage to the spinning logo.

Premiere Pro produced jagged edges and other artifacts in most clips, and severe tearing of the spinning logo. Adobe very reasonably responded that since the price of the Adobe Video Collection was $799 (at press time), less than some of the products here, and included Adobe After Effects Standard, that we should also test the After Effects keying capabilities.

We experimented a bit with the After Effects filters, and got improved results, but couldn't produce comparative clips in time for publication. Adobe also noted that they were reworking their DV codec to produce better keys, an enhancement to be available in their next release. Until then, users serious about chroma keying should probably count on having to acquire After Effects for high-quality work.

(Click Figure 3 to see a full resolution sample comparison of color keying from our tests.)

Color Correction
What's remarkable about color correction is that often you don't know how bad you need it until you get it and try it. All of the programs provide extensive color correction controls, with near infinite flexibility. Unlike our keying trials, where true qualitative differences exist, all products can probably produce identically color-corrected files, if the operator is given sufficient time. In this regard, the quality rating is really more of an ease of use indicator than a measure of absolute quality. To test this, we used four clips, two discussed here, both victims of inadequate white balancing. The first, shot at Zoo Atlanta, had an overly bluish tinge, while the second, shot at a wedding, was way too rose-colored (even for a new couple).

Our winner in this category, Pinnacle Liquid Edition, provides the best range of tools in the review. The product's Automatic Color Correction utility is a model of simplicity, with one eyedropper and four adjustments. Drag the eyedropper and touch a white spot in the video, and Edition automatically white balances the video, which you can then adjust. Unlike most other products, which offered seemingly dozens of dials and levers, having only four controls was a relief, proof that less is often more when it comes to tool design.

(Click Figure 4 to see a full resolution image of the correction color of our comparison sample.)

TABLE 2: Color correction features, performance, and quality rating

Not only was the Automatic Color Correction best in terms of quality, it was also the fastest, with much less time invested than with any of the other four products. As shown in Table 2 (above), Edition also offers a very wide range of scopes and other tools, providing a great balance of tools for both novice and expert users.

Next up are Sony Vegas and Apple Final Cut Pro. Both companies offer visual tools with three wheels for low (blacks), midtone, and high (whites) colors, with eyedroppers to choose representative colors in all categories, similar to the Edition approach. Vegas, however, tended toward over-saturation in testing, producing videos that looked like colorized versions of black and white films, though this was fairly easy to compensate for. Vegas does provide a split-screen preview and a range of scopes.

In contrast, Final Cut Pro's automatic adjustments proved insufficient, forcing lots of manual adjustments. Fortunately, most could be made in the program's "visual" mode, because switching over into "numeric" mode plunged you into a complex series of adjustments with unfamiliar terms like saturation width, minimum, and softness. Seasoned pros will master these adjustments easily, but those moving up from iMovie3 or a Windows-based consumer NLE will find them arcane. In short, achieving good results in Final Cut Pro was time-consuming.

Avid Xpress Pro was guilty of being too good at automatic adjustments, providing the simplest one-button adjustment in the review. The problem was, the correction looked great until you compared it to the output of other products, which generally removed more of the incorrect tint. You can see this in both the zoo sequence (too much blue) and the wedding sequence (too much pink).

(See Figure 5 to see a full screen sample comparison of our color correction results.)

Though we tried to correct for this, we found the manual controls difficult to operate and often found that the image preview, even after rendering on the timeline, didn't provide an accurate representation of the final rendered image. This, the lack of key frames, plus the need to correct for the slight fading mentioned earlier, complicated the correction process. Interestingly, however, without other clips to compare to, Xpress looked great, curing about 90% of the problem with one simple click.

Premiere Pro launched a kitchen-sink control that includes color wheels, histograms and vector scopes, and simpler tools, like the Black, Gray, and White Point adjustments that work very much like Edition's auto correction tool—you touch an area that should be black, gray, or white, and Premiere makes it so, magically adjusting all other colors. Of course, you can use all other tools to fine-tune the results.

In practice, Adobe's control was much more "twitchy" and tended to produce bizarre, almost psychedelic results if you selected an area on screen that Premiere didn't like. This was easy enough to correct (just choose an adjacent area), but definitely slowed the process.

(See Figure 6 for more full screen color correction comparisons.)

Strangely, if you render your file with the split-screen preview enabled, Premiere renders the file with one side color-corrected and the other not; if this sounds like a gotcha that can getcha, it is, and we re-rendered several times to correct our split screen videos. Most importantly, corrected quality simply lagged that of most other products.

Once again, treat this section more as an ease-of-use issue rather than an absolute measure of capability. A long-time user of any of the five programs, especially one familiar with vectorscopes and similar tools, could likely equal or exceed Pinnacle's quality in short order. Less experienced users, however, will likely find greater success with Edition and Vegas, primarily due to the simplicity of their tools.

TABLE 3: Slow-motion features and ratings

Slow Motion
Slow-motion treatments are great for showing action and for dramatic effect. To test this capability in each product, we used four test clips, two discussed here. The first was a chip shot slowed to 25% speed, the second a full-speed golf swing we attempted to slow to 50% and then 25% speed gradually within the clip, and then gradually speed back up to 100%.

Success in this category relates primarily to feature depth, with two focal points. The initial focus is how the editor creates the additional frames that enable the slow motion. There are two basic techniques for this, blending and duplication.

During blending, the editor looks at the actual video fields and creates "tweener" frames that simulate where the object in motion should be for those particular frames. This works well in many cases, but where detail is high, as in our slow golf chip shot, artifacts like multiple clubs appearing in the same frame can occur.

In contrast, duplication merely duplicates the frame as necessary to fill the required frames. For example, slowing video speed to 25% requires three frames for each original frame. Frame duplicators simply duplicate the original frame three times, producing a crisp image that is choppy in appearance.

In between these two techniques are other interpolated styles that attempt to retain crispness while smoothing motion. We'll discuss these techniques with the products that offer them.

The second feature we analyzed was the ability to vary speed smoothly over time. Sometimes, it's acceptable simply to cut to a clip running at a different speed, while other times it looks better when you smoothly shift speeds, for example, taking one second to transition from full speed to half speed.

In addition to these feature-based comparisons, we also analyzed quality, looking for artifacts created when producing the slow-motion effect, which differed from product to product.

As shown in Table 3, Edition offers both the most techniques for slowing speed, and a reasonably easy-to-use control for variable-speed changes. This, plus the lack of significant artifacts in the video, resulted in the top score in this category.

Avid Xpress Pro's VTR-style technique for creating tweener frames produced the best motion quality that we saw, but at a slight cost of background shimmering. This, plus the lack of customizable variable-speed effects, cost it a perfect rating.

Vegas offers only an interpolated style of frame replication, which produced multiple golf clubs in our test clip. On the plus side, Vegas has the simplest variable speed control we've seen (and one we wish other companies would adopt): You insert a speed envelope, which introduces a rubber band control you touch to create a keyframe and then adjust up and down to moderate the speed. Unless you're working with high detail clips like our chip shot, Vegas should work very well.

Final Cut Pro offers both blending and duplicated style frame creation, with very limited artifacts. The limiting factor here was the ease of use of the variable speed tools. Though there were several alternatives, we found each very complicated and unintuitive, especially compared to Vegas.

(See Figure 7 for a sample comparison of artifacting in our slow motion test results.)

Premiere Pro uses a blended style, producing multiple clubs in our chip shot, though quality was good in all other test clips. Premiere does not have a dynamic motion feature, so you can't gradually change speed. More troublesome was a very noticeable shimmering on the brick wall behind the chip shot, which Adobe attributed to "a limitation in the current algorithm."

(Here is an example of variable speed adjustment in Vegas and Final Cut.)

Image Stabilization 
Unless you park your camera on a tripod, image stabilization--removing the jitters and shakes from your footage--is a constant problem. Three of the five products reviewed here offer solutions--Premier Pro, Final Cut Pro, and Xpress Pro.

We tested image stabilization with three scenes. The first two, a video of a parade shot while walking alngside the participants and the second, a video shot of a tour guide in a moving bus, had verry jarring motion. The third was a handheld camera shooting a concert without a tripod, with small but noticeable motion.

Of the three, we found Premiere's plug-in Steady-Move module from 2d3 to be the most functional and easy to use. Simply applying the filter using the default parameters reduced motion significantly on the first two clips and tended to slow the motion on the third clip, making the jitters much more palatable, with no noticeable side effects. The result was impressive enough to consider using for all handheld footage.

With both Xpress and Final Cut, you operate the filter by selecting a region in the video that doesn't move during the course of the shot. This wasn't possible in the parade video, but we selected a spotlight on the bus and a light fixture in the concert video.

With this direction, Xpress reduced the shakes to some degree, but not nearly as dramatically as Premiere. Final Cut did the same through some stretches, but often jerked the video violently, degrading quality unacceptably. We recommend caution when using this filter.

Conclusion
Apropos of the elephant metaphor that opened this article, we understand that this article barely touches upon the enormous feature sets of each product, and ignores many of each product's key strengths. Still, however limited, the functions analyzed here represent critical, make-or-break capabilities of each editor.

TABLE 4: Image stabilization (where offered)

COMPANIES MENTIONED IN THIS ARTICLE
Adobe Systems, Inc. www.adobe.com
Apple Computer, Inc. www.apple.com
Avid Technology www.avid.com
Pinnacle Systems, Inc. www.pinnaclesys.com
Sony Pictures Digital www.mediasoftware.sonypictures.com