Ogg Vorbis Review

When last I visited Ogg Vorbis ( http://www.xiph.org/ogg/vorbis/), the unlikely-named, free-as-a-bird audio CODEC was in second beta and being distributed under the LGPL, or Gnu Lesser General Public License. The LGPL basically says that anyone using the freely available Ogg source code must return any modifications to the general public — open source at its most benevolent. Although a slower encoder than Fraunhofer's MP3 or LAME (a free MP3-like CODEC), Ogg Vorbis beta 2 produced files of like size and even better fidelity.

Now in fourth beta, Ogg has been re-released under the FreeBSD license that keeps it free to the public and open source forces, but allows companies to retain the rights to their proprietary modifications — a scheme which will hopefully stir more interest from commercial sources. Jack Moffit, the Grand Vizier of Vorbis told me that beta 4 both sounds better and produces smaller files than beta 2. I believed him, but of course I had to see for myself.

Already having tested Ogg beta 2 against Fraunhofer and LAME, I decided to go a bit more mainstream (puns intended) in my tests this go round and pitted Ogg beta 4 against Windows Media Audio, RealMedia 8.5, and AAC (Liquid Audio 5.0) as well as the MP3 CODECs. I used 128-Kbps encoding, or as close as possible. I then compared the size of the resulting files and performed listening tests on a Pioneer VSX-454 receiver driving Boston Acoustics CR-7 bookshelf speakers. My test file was 'Just a Girl' by No Doubt, which features a wide variety of dynamics, a relatively dry mix, and an easily distinguishable bass tone. The times it took to encode under Windows 98 on a 1.2GHz Athlon system and the resulting file sizes for each CODEC are listed below:

CODEC bit rate file size time Ogg Beta 4 128 kbps VBR 3,293,184 0:43 Lame MP3 128 kbps 3,334,144 0:31 Ogg Beta 2 128 kbps VBR 3,514,368 1:13 Real Audio 8.5 132.2 kbps 3,530,752 0:14 Fraunhofer MP3 (Sound Forge 5.0) 128 kbps VBR 3,858,432 0:09 Windows Media 7 (Sound Forge 5.0) 128 kbps 3,383,296 0:09 AAC (Liquifier Pro 5) 128 kbps 4,210,688 0:57

Ogg was the slowest encoder other than the Liquifier Pro 5 (AAC), but it also produced the smallest file. The fidelity of the compressed songs was generally excellent. However, while RA, AAC, Ogg, and WMA were all difficult to distinguish from the original wave file, the Fraunhofer and LAME MP3 both noticeably (though not annoyingly) exaggerated midrange frequencies. The size of the Liquid Audio/AAC file is inflated due to security anddelivery overhead. Alas, truth be told, I had a hard time telling the Ogg beta 4 file from the Ogg beta 2 file — both sounded excellent. But Ogg beta 4 is indeed faster and produces smaller files than its predecessor.

Since I lacked the facilities to test Ogg's streaming performance, I instead opted to test how many CPU cycles it consumed. Sadly, a complete apples to apples comparison was impossible since no one player supported all the CODECs. However, Winamp utilized 2 to 3 percent of the available CPU cycles playing MP3, 5 to 6 percent playing RA and WMA, and about 6 to 7 percent playing Ogg beta 4 files. The variance could mean many things; however at the very least I can conclude that Winamp's Ogg plug-in isn't as efficient as the others.

Now I've once again proven that Ogg sounds as good as other CODECs and produces even smaller files, how does this affect your life more than the price of tea in China? Well, obviously it's the lack of bitstream royalties or licensing fees. Exactly how long do you think a pay CODEC would stay reasonably priced if it succeeded in becoming the de facto standard? Capitalism is not an altruistic religion. If WMA, Real, AAC or another pay CODEC gain market dominance, kiss low-budget streaming operations good-bye.

Okay, so I sound like a true believer; I also run Linux and recycle, what can I say? The bottom line is that Ogg Vorbis is free and it works. Here's hoping the Vorbis project can stay afloat long enough for it to gain some popular support.