In addition to boxed and download versions, Adobe is introducing subscription pricing for users who want to subscribe on a monthly or annual basis but just pay monthly, rather than purchasing the suite outright. Subscription users can select individual programs or the entire suite, and depending on their contract, they can keep them activated for as little as a month. I can see this being very popular for studios that want to install CS5.5 applications on additional editing systems without the upfront investment, or those who want to activate more advanced programs only when the need arises. Monthly pricing for a 1-year commitment and month-to-month pricing on individual point products ranges from $19-$29 for Dreamweaver to $49-$75 for Photoshop Extended or After Effects. Prices also range from $85-$129 for the full Production Premium suite to $129-$195 for the entire Master Collection. One advantage of choosing the subscription option is that subscribers get automatic updates to the latest version, and that is included with their monthly rates. But enough with the strategy and pricing—let's have a look at the new point products.
Premiere Pro CS5.5
Perhaps the single most important feature of CS5 Production Premium was its new 64-bit Mercury Playback Engine, which delivered incredible GPU acceleration via a handful of supported NVIDIA CUDA cards. In version CS5.5, Adobe has further increased the number of supported NVIDIA CUDA cards to 20. This is good news all around if you are building a new editing system or upgrading an existing one, as it gives you more card options.
For this review, I'm testing two builds. The first is an upgrade from CS5 to CS5.5 on a system that I built in 2010, a first-generation Intel Core i7 920 on Asus P6T motherboard with a GeForce GTX 285 video card. I'm also testing a full new install on my latest build, a second-generation Intel Core i7 2600K on Asus P8P67 Deluxe motherboard with a NVIDIA Quadro FX4800 workstation graphics card. My upgrade process proved uneventful, which is a good thing, although I had to update my graphics driver in order to unlock the GPU acceleration. I probably should have taken this as a hint of additional performance improvements but, as I'll explain later, I didn't think much of it at the time.
Now, I really shouldn't call Premiere Pro CS5.5 an upgrade, as it's a full stand-alone version and CS5 was still available for me to use (much as CS4 was still available when I installed CS5 last year). Of more importance, because both CS5 and CS5.5 were available on the same system, I was able to easily test for performance improvements, if any. Because encoding and exporting tasks in Premiere Pro are handled by the stand-alone 64 bit Adobe Media Encoder, I'll discuss the improvements further on. So be sure not to skip that section, as it reveals some shocking results.
As part of the advance review process, Adobe provided me with a reviewer's guide that outlined updates to the individual point products (as Adobe calls its individual applications). Several of the new features in CS5.5 didn't seem that new to me, and when I checked, they were only newish, as they were also available in CS5. So I'll focus on a few of the authentically new features here, which include the new GPU-accelerated, real-time film dissolve that uses linear color blending to prevent edge and halo artifacts, such as fringing. Fringing can occur when high-contrast saturated colors are blended together. GPU support is also now available to accelerate several additional and previously available effects, such as blurs and the additive dissolve.
It's worth noting that Adobe Story, the script-writing program that Adobe launched in CS5, has also been updated. However, because it isn't as relevant to the corporate and event market as it is to filmmakers, I didn't review it. I also didn't review the closed captioning support, but I plan to at a later date, as I occasionally get demand for that feature. Premiere Pro can now import data in 608 and 708 formats. When used in conjunction with the improved speech analysis function, this feature allows editors to navigate based on words and creates metadata that, because of the tight integration of Adobe products, follows the video right through to its delivery format.
For me, by far the most exciting new feature in Premiere Pro CS5.5 is the new Merged Clips feature. At first, I thought this would be exciting only for DSLR shooters (and others who work with external recording devices), as it allows the editor to link multiple audio tracks with a video track. This is essential for working with DSLR footage, because DSLRs lack professional internal recording capabilities. But when I saw how Adobe linked the two media, I realized that the feature would also benefit multicam shooters working with traditional video cameras as well.
The new Merged Clips feature in Adobe Premiere Pro CS5.5
When you link a video track with one or more audio files, Premiere Pro creates a new linked file in the Project panel (which you can name), without any additional rendering or hard drive storage requirements. The new linked file acts like a regular linked video and audio file, and the original unlinked files can be archived in a bin or deleted from the project folder without deleting the new linked reference file. This is different from simply nesting a sequence; a nested sequence lacks the useful audio waveform display, and the original assets need to be preserved in the project window.
Video producers who frequently edit multiple camera angles on live events know that it's very common to split the responsibilities of recording the individual isolated audio feeds among the camera operators. On a basic two-camera shoot for a dance recital, this would mean that one camera operator would take a soundboard feed on one video camera's XLR inputs and run a shotgun microphone for ambient on the second. The other camera operator would then record the isolated stage microphone on one of her camera's inputs and possibly an isolated announcement and intro microphone feed on the second. When all of these feeds are dumped on a timeline, each of the two stereo audio feeds needs to be duplicated. Then, one audio channel needs to be filled to the right and the other filled to the left before the four audio inputs can then be mixed in preparation for the delivery format, which will either be a single stereo or mono audio track.
Does this approach sound inordinately confusing and complex? Even more complicated is that, with the old method, the fill channel effect doesn't update the waveform, so both the duplicated and cross-filled audio channels appear identical. If you want to use the peak amplitude reading on the audio gain function to raise the level of a track, it gives a reading for the peak of the loudest of the original audio tracks, even if you already filled one track to the other. It's possible to render and replace each audio track, but that takes time and hard drive space.
Premiere Pro's new Merged Clip feature does more than just link multiple audio tracks to a video track. To my pleasant surprise, when I dropped the merged clip reference file onto a timeline, the pair of stereo tracks dropped in as four individual mono audio tracks. Having individual isolated audio tracks, as opposed to a pair of mono tracks recorded as a stereo file, solves all of the problems attendant to the old CS5 workflow. The waveforms are visible for each track, the peak amplitude reading reflects the individual isolated audio track, and there is no more duplicating and filling.
Adobe Media Encoder CS5.5
Media Encoder, Adobe's stand-alone video encoder, received a makeover in CS5.5. Visually, the Watch Folders function gets promoted from being hidden in CS5 to being on the main interface in CS5.5. Though it's not a new feature per se, increased prominence brings much needed attention to this useful feature. Once a watch folder is created, editors can easily set it up to automatically create additional codec version encodes for every file they create or drop into the watched folder. This is useful for projects that require delivery across multiple formats, including the new presets that were created for tablets, mobile devices, desktops, and television.
Working with watch folders in Adobe Media Encoder CS5.5
The good part of this workflow is that, if done properly, it automatically creates a master copy for you to archive. The bad part is that your delivery encodes will be rendered from the master and not the timeline. Purists will argue that, theoretically, this is a step removed from the source, and the master is a second-generation file. To minimize potential theoretical generational loss, I recommend exporting master files by checking the Match Sequence Setting box instead of trying to match the codec yourself. I wish the entire watch folder process was one step easier and closer to a pure first-generation encode, but I tried dragging a Premiere Pro CS5.5 sequence directly into the watch folder and it wouldn't take. So add that to my wish list, Adobe, along with improvements to the preset export settings.
In my 2010 review of Adobe CS5 (http://bit.ly/Lam-Adobe-CS5), I ranted about the confusing and often incorrect frame rate settings that afflicted Adobe's export presets. The problem was that Adobe assigned frame rates in the presets to a seemingly random sampling of both drop and nondrop frame rates. Rendering a video away from either the source or sequence frame rate is a bad idea. Not only does it add significantly to the encode time (in one of my tests, it doubled the encode time), but it can also lead to some very undesirable effects when frames are being dropped or repeated. Unfortunately, Adobe has done nothing to correct this issue with CS5.5. Maybe we'll see an update in CS6. Rather than creating multiple presets for each possible frame rate, I'd like to see Adobe create a little check box called Match Frame Rate that allows users to modify the presets (that are fine otherwise) with the timeline frame rate. Adobe has already built in a Match Sequence Settings check box in the applications, so I don't think that's too much to ask.
Adobe Media Encoder CS5.5, like its CS5 predecessor, has a helpful Match Sequence check box for matching export settings to the sequence, but no check box for Match Frame Rate.
Now, after participating in an advanced conference call, reading every page of my reviewer notes, and installing and using the final build of the CS5.5 Production Premium, I focused all of my initial review attention on new features. It wasn't until a few hours before my already-extended editorial deadline that I thought of testing CS5.5 for speed improvements. I admit that I wasn't expecting any. The reason is that neither my conference call nor my reviewer notes made any speed improvement claims, outside of a few new accelerated effects. So when I tested same-system performance against CS5, I was absolutely blown away by CS5.5's dramatic and unexpected performance improvements.
My current PC build
Much as in my CS5 review, I created two HD sequences, this time with XDCAM 1080 30p footage. The first timeline was 60 seconds of straight video with no effects. The second was 10 layers of video, including five HD video tracks, one SD video track, two logos in Illustrator format, and two white color mattes. I also added a color correction, four PiP scaling effects, some cropping, and some opacity adjustments. Needless to say, this turned out to be a pretty complex 60-second sequence and not something that I would likely ever create in a normal workflow. But the point was to push Premiere Pro's limits and see what happened.
On my year-old, first-generation Core i7 system, CS5 exported the single-layer timeline in 0:59 (minutes:seconds) and the 10-layer timeline in 3:22. As I wasn't expecting any substantial improvements over CS5--especially given that none of my timelines incorporated any of the newly accelerated effects that were noted in the reviewer's guide-I was shocked with the improvement that I saw with the CS5.5 render times on the same system. The CS5.5 results were 0:33 and 1:25-a 44% improvement on the single-layer render and a 58% speed jump on the 10-layer.
I can attribute some of the improvements to better use of the CUDA cores on NVIDIA's graphics cards, which offload some of the video processing from the computer's CPU to the parallel processing video card GPU. My Windows Task Manager confirmed the drop in CPU use.
When I was satisfied with my results that CS5.5 was faster than CS5, and in the interest in time, I almost skipped testing the same exports on my new system build (a second-generation Core i7 system). But curiosity got the better of me, and I continued the testing. The results were confusing. My single-layer time dropped from 0:33 to 0:24, but the 10-layer time increased from 1:25 to 2:06. This meant further benchmarking was required. In my CS5 testing, I concluded that it didn't make any difference which Adobe-approved NVIDIA CUDA card you used, because the results were all the same. So in my CS5.5 tests, I didn't initially bother to make sure I was using the same graphics card on both systems. In search of an explanation as to why my new build was providing slower results, I removed the ASUS GTX285 (a performance gaming card shipped in 2009) from my first-generation i7 system and installed both it and new drivers in my second-generation system. That system had previously housed a NVIDIA Quadro FX4800 workstation card. I reran the exports. My export time on the 10-layer sequence with the GTX285 on the new second-generation Core i7 build matched the 1:24 time with the same card in the older build.
As far as I can tell, the major differences between the two cards that I tested are that the FX4800 has 192 CUDA cores and draws only 150 watts, while the GTX285 has 240 cores and draws 249 watts. The cost is also a major difference but the performance-to-price ratio is the opposite of what you would expect. The GTX285 was listed at only $399 in 2009, although it was already discontinued when CS5 launched. I bought mine used for $200, while the FX4800 was listed at $1,500 in 2010. I asked Adobe to confirm my findings and the following statement: "When using an approved NVIDIA card, CS5.5 performs better using cards with more CUDA cores." A few days later I received a short but to-the-point email confirming that my statement is accurate.
Just in case you're wondering, in addition to eight models within the workstation Quadro line, and five mobile workstation Quadro models, CS5.5 will support the following GeForce consumer desktop graphics cards (prices quotes are from online retailers in April 2011):
• GTX 580, 512 cores, $500
• GTX 570, 480 cores, $330
• GTX 470, 448 cores, $250
• GTX 285, 240 cores (2009 model, unavailable)
So for 2011 and CS5.5, I'm going to change my CS5 recommendation of buying the least expensive graphics card you can find to buying a GTX 470 or 570, as I expect them to deliver the best performance-to-price ratio.
Sacrificing Image Stability for Bokeh
When video historians look back at 2011, they will probably shake their heads in disbelief at how many video production studios no longer own and operate video cameras. The rise of the video-capable DSLR can be traced back to 2009, but by late 2010, DSLRs reigned unchallenged by video cameras in producing the sought-after bokeh defocus. Largely, this is because achieving defocus requires a much larger sensor than the 1/3" or 1/2" varieties to which sub-$10,000 video cameras were limited. Fortunately for the future sales prospects of video cameras, DSLRs have two problems: They lack XLR audio inputs, and their form factor makes it difficult to get steady shots when operated handheld. Manufactures have rushed to fill the need for DSLR rigs, and producers with larger pockets can afford an arsenal of image-stabilized lenses, but many producers still need image stabilization software when working with DSLR footage.
Part of the solution comes from Panasonic, the first company out of the gate to offer a video camera with a DSLR-quality imager, the APSC sensor-based AF100. Sony followed with prototypes and a promise of a Q2 2011 release of its Super 35mm NEX FS100. Both of them share traditional camcorder form factors, which are easier to hold steady than a DSLR, but many of the lens options lack optical image stabilization. Regardless of which format will reign supreme, the trend is clear: Video cameras of the future will feature large sensors and interchangeable lenses. And because of the interchangeability of the lenses, image stabilization will shift to a problem that increasingly will be solved in post, and CS5.5 Production Premium introduces a great image stabilization solution.
Adobe After Effects CS5.5
I don't normally use After Effects in my studio on a daily basis. My editors and I usually reserve it for problem solving and complicated graphics work. This is about to change because of one feature, which alone is worth the upgrade price: the new Warp Stabilizer. Not only is this feature easy to use, even for non-After Effects users, but it works much better than anything I have ever seen for camera stabilization.
In true Adobe fashion, the Warp Stabilizer can be accessed in at least two ways. The first is from the drop-down menus (Animate then Stabilize) or by searching "Warp S" in the Search. In my recent EventDVLive feature Producing Conference Video (http://bit.ly/Lam-EDVLive-4), I hurriedly shot and edited an accompanying video moments before my client was set to arrive. Halfway through filming a handheld sequence, I realized that I hadn't engaged my video camera's optical image stabilizer, which I always turn off when shooting on a tripod. I didn't have any time to reshoot, so embarrassingly, the shaky footage made it into the finished product. Now Adobe has given me the tools to redeem myself, as that shaky video served as some great test footage for me to use in After Effects CS5.5 to test out the Warp Stabilizer.
After Effects CS5.5's Warp Stabilizer feature analyzing my handheld footage prior to stabilization
The default settings produced great results, but I decided to delve into the custom settings to see how much control I could have in case the default settings didn't stabilize my video or introduced undesired effects. The first option is to choose between Smooth Motion and No Motion. Smooth Motion is the default and, as the name implies, it stabilizes the video but leaves some camera motion; No Motion, by contrast, locks the image as if it was shot on a tripod. Then there are four method options, including the powerful Subspace Warp, and finally Framing. The Framing option controls the degree of scaling from none to some. If you select no scaling, a black border will appear, showing how much data was discarded in order to get a stable core image. If you don't want to scale, or have a black border, you can select Synthesize Edges and After Effects CS5.5 will fill the borders using information from adjacent frames. Be sure to watch the accompanying video for split screen examples of three different stabilization options and a sample of the new After Effects CS5.5 Camera Lens Blur feature.
Step 2: stabilizing the footage
Adobe, with its CS5.5 improvements in speed, postproduction fixes, and real-time creative effects, has done its part to free video editors worldwide from countless hours of isolation from the edit suite. Not only will the speed improve turnaround times, it will also provide editors with the luxury of experimenting, which is an important part of the creative process. Now, if only they'd fixed the frame rate preset problem ... well, I suppose they have to leave us something to wish for in CS6.
Shawn Lam (video at shawnlam.ca) runs Shawn Lam Video, a Vancouver video production studio. He specializes in stage event and corporate video production and has presented seminars at WEVA Expo 2005–9 and the Video 07. He won Creative Excellence Awards at WEVA 2010 and 2008 and an Emerald Artistic Achievement Award at Video 08.