2009 Encoder Shootout
Wow! Another year has passed. And so it’s time for another round of encoding tool updates. In this roundup, I’ll compare the output quality, speed, and codec-related feature set of the most prominent crop of sub-$1,000 encoders and provide the results in report-card form. For the most part, the best in each category rated an A, the worst a C, and the rest B’s. Though this will vary by the encoding tool and category, a C grade should be a concern if you’re currently using or plan to use the encoding tool. Where a feature is significantly limited in a particular encoder, I've noted it in the tables with asterisks and explained the issue in the text.
Briefly, for quality tests, I produced SD (640x480 at 30 fps at 468Kbps video/32Kbps audio) and HD (1280x720 at 29.97 fps at 800Kbps video/128Kbps audio) files using H.264, VP6, and VC-1 codecs; I compared output quality with other encoding tools using the same operating system. I produced all comparative quality tests from prescaled, predeinterlaced files to isolate the encoding quality rather than the scaling and deinterlacing quality of the respective tools. To test scaling and deinterlacing quality, I used a 1-minute DV file containing multiple scenes with sharp edges, diagonal lines, and other jaggy magnets; I then compared the output of the respective tools.
To test performance, I timed the encoding of a single 1-minute file and then timed how long it took to encode five 1-minute files, deploying all available techniques to speed encoding time, such as running multiple instances of On2 Flix Pro and Sorenson Squeeze on both the Mac and Windows platforms for the five-file encoding trials. For Windows, I tested using a 2.8-GHz Dual-Processor, Quad-Core HP xw6600 workstation running XP with 3MB of RAM. For the Mac encodes, I tested on a 3.2-GHz Dual-Processor, Quad-Core Mac Pro running Leopard with 8MB of RAM.
In terms of supported features, this varied by codec. For H.264, I looked at features such as access to the baseline, main, and high profiles; support for context-adaptive binary arithmetic coding (CABAC) entropy encoding; and the breadth of B-frame configuration options. For VP6, I looked at access to two-pass encoding and also the VP6-S and VP6-E options. For VC-1, I looked at access to the tweaking parameters enabled in WMV SDK 11 and support for Silverlight output.
Adobe Media Encoder CS4
CS4 was a great release for the Adobe Media Encoder (AME), as the program gained both batch capabilities and stand-alone operation, meaning that you can run the program separately from Premiere Pro (or other CS4 apps) and insert files or Premiere Pro sequences for encoding. Even more impressive, you can send sequences from Premiere Pro to AME and continue editing in Premiere Pro, a dream feature that easily justifies the suite upgrade price for most serious Premiere Pro editors.
There are some limitations, however. For example, the new watch folder functionality doesn’t trigger automatically when a file gets dropped into the folder; you have to start the encoding process manually, a limitation that hopefully will be addressed in a future release. The other negative is that you can’t open multiple instances of AME to boost the speed of multiple file encodes, which you can do with many other programs, including Microsoft Expression Encoder 2, On2 Flix Pro, and Sorenson Squeeze—the latter two on both Mac and Windows. Nor does Adobe offer a tool like Apple’s Qmaster to make encoding on multiple core computers like our 8-core HP xw6600 more efficient. Since AME is very fast with Windows Media and H.264 output, this primarily impacts VP6 encoding times, as I detail below.
In terms of output quality, it was a different story for the Mac and Windows versions, with Windows offering better H.264 quality and the ability to produce WMV files, which AME can’t do on the Mac. On both platforms, H.264 configurability is limited. You can choose the baseline, main, and high profiles, but you can’t choose CABAC entropy encoding or control B-frame sequencing. Not surprisingly, though, you can output in the new F4V format for Flash.
Still, in Windows, H.264 quality was quite good—very close to that produced by Sorenson Squeeze, which produced the best quality among the reviewed programs. Mac H.264 quality was spotty, though—poor in our SD trials and very good in HD. AME was a screamer in our performance trials, with the fastest five-file encoding times on both the Mac and Windows. Overall, if you’re a casual H.264 producer, AME is fine in Windows, but you may wish to look elsewhere for the Mac.
The big story with VP6 is that AME gained two-pass encoding capabilities, which boosted output quality to the top tier, and access to VP6-S, the "simple" version of the codec that’s ideal for HD video or video produced for low-power devices. As mentioned before, the big negative for AME and Flash is encoding time. On our 8-core HP workstation, for example, Squeeze produced five 1-minute test files in 10:28 (min:sec), while it took AME CS4 33:30 to accomplish the same task. The difference was even more stark on our 8-core Apple loaner: On2 Flix Pro took 6:22; AME took 47:20. If you’re trying to encode Flash files in a hurry, AME is not for you on either platform. If you’re not in a hurry, it’s a great choice.
On the Windows platform, Windows Media quality is good and encoding times are quite good; AME should prove adequate for any producers not looking for the ability to tweak their Windows Media files or create a Silverlight player.