Video
Compression (Part 1 of 4)
Using Final Cut Pro and the Sorenson Video Codec
to Prepare QuickTime Video for Web Distribution
My next-door
neighbor, Jordan Rudess, is a pianist, synthesist, and composer who has
recorded and performed with many well-known artists, including David Bowie,
Tony Williams, Enrique Iglesias, the Dixie Dregs, the Paul Winter Consort,
and Liquid Tension Experiment. Jordan is a virtuoso of high order who
entered Julliard at the age of 9 to study classical piano, but was seduced
by progressive rock along the way; he's currently a member of Dream Theater,
who've just released a new album and will be embarking on a world tour
in January. In addition to all this, Jordan teaches privately and runs
his own website (http://www.jordanrudess.com),
where he posts updates on his projects, interacts with his fans, and sells
his recordings and instructional videos.
Recently, he told me that he's been getting requests from potential students,
mostly from overseas, to deliver his training videos on the Web. He looked
into converting some of his analog videos into RealVideo format and streaming
them from his server, but they sounded terrible, and the image quality
and frame rates were so low that you couldn't see what he was doing, rendering
his demonstrations of fingerings and other keyboard techniques useless.
I suggested that he try downloadable QuickTime movies using the Sorenson
and QDesign codecs instead, which would provide a much higher image and
sound quality, and offer the ability to lock the video track before sale
to help limit piracy. Since I have a consumer DV camcorder, the Canon
Optura, the Developer version of the Sorenson Video codec, and a G3 PowerBook
DV editing system (including a Ratoc FireWire card and Apple's Final Cut
Pro), I offered to shoot Jordan in his studio myself, edit the footage
in Final Cut, and Sorenson-compress it for Web delivery with Terran Interactive's
Media Cleaner Pro, all on my PowerBook.
You'll see the term "data rate" a lot in this article. Data rate simply
means how many kilobits, kilobytes, megabits, or megabytes per second
is required for realtime playback of a given video or audio clip. For
purposes of comparison, here are some standard data rates: Uncompressed
broadcast-quality digital video has a data rate of 18 megabytes (MB) per
second. This means that in order to see this type of digital video in
realtime at full frame size and without dropping frames, the system playing
back the video needs to be able to deliver a steady stream of data at
18 MB/sec - a data rate few standard PCs are able to sustain. Standard
DV video, on the other hand, being somewhat compressed, has a data rate
of 3.5 MB/sec, which is more reasonable for current PCs. Standard CD-quality
audio (44.1 KHz/16 bit/stereo) has a data rate of 150 kilobytes (KB) per
second. This is also the delivery speed of a 1x CD-ROM drive.
Compare that to the data rates an Internet connection is able to provide.
A 56K modem can deliver a theoretical maximum data rate of 56 kilobits
(not kilobytes) per second; this translates to 7 KB/sec (8 kilobits equals
1 kilobyte). A full ISDN connection of 128 kilobits/sec provides a data
rate of 16 KB/sec. This means that for a video clip to stream in realtime
without downloading via an ISDN line, the clip cannot have a data rate
of more than 16 KB/sec. This is less than 1/1000th of the uncompressed
broadcast video data rate of 18 MB/sec, which corresponds to a compression
ratio of more than 1000 to 1. Even compared to the already-compressed
DV data rate, 16 KB/sec video has a compression ratio of more than 200:1.
This is a lot of compression, and inevitably the sound and image will
be significantly degraded. You can reduce the degradation by reducing
the frame rate, so that fewer frame per second go by, and reducing the
frame size, so there are fewer pixels to compress, but you can only take
that so far before you can't see the video image anymore.