QUOTE (phong @ Jun 3 2004, 09:37 PM)
The VBR setting used for the 128k test recently (-V5 --athaa-sensitivity 1) did very well, so I suspect VBR is "vindicated" at that bitrate, probably lower.
I'm not so enthousiastic. The -V 5 preset performed very well on most samples used in the last test, but these samples are “difficult” to encode. They are maybe common, but they're still difficult.
I've recently encoded some albums in -V 5. Classical music, with wide dynamic, near-silent part, very quiet moments, etc... Bitrate sometimes drop to worrying level (70-80 kbps). Quality is in consequence: bad. Sometimes awful after MP3gain adjustment (I've encoded these albums for my portable).
Audiobooks, with speech, are not comparable to the samples tested recently. I suppose that ABR is the best solution, or at least the safest.
EDIT: the Debussy sample is close to be a speech sample. Lame VBR was unimpressive, clearly worse than ATRAC3 and probably worse than WMA: