On Thu, 1 Nov 2001, Ross Vandegrift wrote: > First of all, if I use the '--ogg' switch to lame, does lame use GPSYCHO > to encode the wave, or some other psychoacoustic model (perhaps one designed for > the vorbis algorith)? Along these lines, I'm unable to get lame-3.89 and > libvorbis rc2 to play nicely together. The lame build breaks when compiling > ogg support (this is probably a known bug, so I'm not posting relevant details. > If I'm wrong about it being know, we'll go from there). >
Ogg encoding in lame has been broken for quite a while. But even when it worked, it was just an interface to the ogg libraries, and produced the same results as oggenc. It was added because originally the ogg encoding tools didn't have features lowpass filterings and downsampling. Now we are too lazy too decide if it should be fixed or removed :-) > Second, since audio compression is based on the idea of a plausable > psychoacoutic model. There seems to be some documentation on GPSYCHO, which > lame uses. But the link 'The Vorbis probability model' on xiph.org is noteably > not a link. Is there any addition documentation about the psychoacoustic model > that vorbis uses? > The only real documentation would be the ISO docs themselves. The two psy-models in lame, gspycho and --nspsytune are both modeled after the ISO psy-model (using ideas from both MP3 and AAC). They differ in things like how tonality is estimated, how pre-echo is detected, but the basic principles are the same. I personally believe that the noise shapping (how you distribute the bits among the coefficients, given a set of masking data) is the dominant determination of quality. I dont know if noise shapping algorithms are considered part of the psycho acoustic model? Mark _______________________________________________ mp3encoder mailing list [EMAIL PROTECTED] http://minnie.tuhs.org/mailman/listinfo/mp3encoder
