The 2008 Encoder Shootout
If you select the QuickTime option, you can hint your video for a streaming server, but you’re limited to 1-pass encoding and can’t set a data rate for audio. Instead, you choose the output channels and frequency, and the Adobe Media Encoder assigns a data rate. This lack of feature support spoils a good effort by the MainConcept H.264 encoder, which produces a very high-quality file only a hair behind Apple Compressor and Sorenson Squeeze in the most challenging scenes.
Semaphore revealed that the Adobe Media Encoder accurately inserted keyframes at scene changes, but also appeared to use a GOP size of 30, inserting a keyframe each second, which is much more frequent than other encoders, and probably contributed to the quality degradation during extreme high motion. However, the data rate was very smooth, peaking out at about 750Kbps, compared to Compressor’s peak of 1.12Mbps.
Flash encoding options are also limited, though less severely. Specifically, Adobe Media Encoder can produce video using only one-pass techniques, where 2-pass is theoretically necessary for optimal quality in the most challenging scenes. However, I did not find this to be true in the Flash file that I reviewed, which looked very similar to that produced by Flix Pro.
Apple Final Cut Pro 6/Compressor 3.02<>
Compressor can batch encode both multiple files in a single project (File A to output X, Y, and Z) and produce multiple jobs to multiple outputs (Files A, B, and C to output X, Y, and Z). You can also create "droplets" for drag-and-drop encoding, a convenience for all users, but especially when setting up repetitive encoding capabilities for nontechnical users. You can assign compression tasks to other computers on your cluster, but there is no watch folder capability. Natively, Compressor supports only H.264 production, though you can produce Windows Media and Flash files via QuickTime Export Components available from Flip4Mac, Adobe, and On2.
The Final Cut Pro-to-Compressor workflow can be confusing; that’s in part an unfortunate byproduct of the product’s flexibility. Like the Adobe Media Encoder, you have multiple export options, including exporting a QuickTime Movie, Using QuickTime Conversion, or Using Compressor, which is generally the preferred option.
In addition, even when you export via Compressor, deinterlacing options vary according to the Final Cut Pro preset that you use, and by compression format. That is, you can’t directly access Compressor’s vaunted Optical Flow technology when producing files via QuickTime Export components, though you can use a feature called "job chaining" to first produce an intermediate file with Optical Flow, then the final compressed file. In addition, the ability to perform operations like converting 16:9 video to 4:3 differs depending upon your output format, largely because QuickTime Export Components don’t use the Geometry pane like native formats.
Even when you do access Optical Flow deinterlacing, it’s often too slow for time-sensitive work, especially when paired with Apple’s high-quality H.264 encoder that requires up to five passes for completion. In one test, for example, producing a 1-minute multiple-pass H.264 file from DV input using Optical Flow deinterlacing took 72 minutes on a Dual Processor, Dual Core 3.0 GHz Mac Pro. Not surprisingly, for this reason, many of Compressor’s own presets use the Better deinterlacing option that does not engage Optical Flow. Unfortunately, there is a noticeable difference between Optical Flow and the next level of deinterlacing, which shows up as jaggies in scenes with lots of sharp detail and diagonal lines.
Since the encoding function is essentially supplied by at least three different vendors, encoding quality depends upon the format. Apple’s own H.264 video codec is top-notch, but seems to have slipped in a recent upgrade. To explain, when I last looked at H.264 quality, Compressor was clearly on top. When I re-encoded clips for this roundup (at least two upgrades later), high-motion quality was way down, and Squeeze had claimed the overall lead.
Comparing the old and new Compressor files in Semaphore, I noticed that maximum quantization jumped from 39 in the older file to 47 in the newer file, indicating significantly greater compression, and that the data rate had stabilized from a peak of 1.12Mbps in the older file to 730Kbps in the newer file, both from our target of 468Kbps. While the smoother data rate of the new file is more manageable from a streaming perspective, this may have also reduced quality during higher-motion sequences.
Companies and Suppliers Mentioned