|
|
|
|
|
|
|
In an effort to help people evaluate the present state of web video codecs, Ocean Street Video has put these clips online. All encoding was done in-house, with the exception of the QT Sorenson Variable Bit Rate video clip. Ben Waggoner of Journeyman Digital stepped up to the plate to showcase the power of the Sorenson codec.
Ben was faced with a less than ideal source file to began with, but all the codecs had the same disadvantage. Some of them handled it better than others, but that isn't a total indictment of any codec/encoder in particular. Think of it more as the disadvantage of comparing codecs based on only one source file. None of these clips use their native server software, all are HTTP streamed.
Given that the visual quality of the non-demo clips is roughly equal, they are ranked according to bitrate/file size. There are two versions of the G2/.asf clips, mostly to demonstrate how low of a bitrate these codecs are capable of. Note that some of these codecs vary the framerate within the clip.
muffmed1sml.asf(demo)......122k
muffmedg2sml.rm(demo).....127k
muffmed1.asf.......................208k
muffmedg2.rm......................238k
muffmed1ben.mov................246k
muffmed1osv.mov................268k
muffmed1.mpg......................341k
The winner in this test may be the Microsoft .asf codec, the picture quality is good and it plays back well. The RealMedia G2 codec appears to have a possible "stutter" problem at the beginning of both clips, apparently because it dropped too many frames there. The G2 low-bitrate demo picture quality appears to be outstanding for the size. However, if there were dropped frames during the encoding process the extra bandwidth would have improved the quality of the rest of the clip. RealNetworks indicates that there may be a "stutter" problem with some (high-bitrate) clips in the player, not the encoder. You can read their comments, as sent to us by senior research scientist Dr. Gary Greenbaum.
The two QT Sorenson clips are rather similar. The larger clip was done in-house, with the $29 version of Quicktime. Playback of these clips on the PC was good. Compare the vehicle color/bleed of both QT3 clips, as well as the area where the fast pan occurs. You can see the effect of the VBR encoding. In this particular case, you may wonder if the ~20% clip size difference(video only) was worth the extra $450+ it cost for the VBR version of the Sorenson encoder. In addition, the VBR clip needs a slightly higher bitrate to match the quality set by the other formats in this test. The Sorenson encoder was, by far, the slowest we tested.
The MPEG 1 clip was encoded by the excellent Ligos encoder, which is perhaps the best software MPEG encoder on the market for web video. While the MPEG file size is the biggest, it is the most universal standard. For example, MPEG 1 is the only web video format that currently plays back on the set-top WebTV decoders. There are over a million of these boxes in use today. MPEG 1 can be somewhat pseudo-streamed with the Intervu Media Player browser plugin.
You can direct all feedback to the email address below, and if there is anything we can do to help you put video on the web, please don't hesitate to ask.
|
|
|
|