lame 3.57beta will soon be on the web page:
(This time i tried to updated the lame.dsp)
CISC: patch to allow encoding from 8bit input files when using LIBSNDFILE
Automatic downsampling to nearest valid samplerate.
Scalefactor bands demarked on MDCT plot in frameanalyzer.
preemphasis disabled. Takehiro's comments have made me doubt the
effectiveness of the current algorithm. Scalefactor preemph
can be turned on, but once it is on outer_loop cant turn it off.
If turning it on is a bad thing, then the only acceptable quantizaton
will be the first one in the iteration with scalefactors all 0.
This way, we get a wider variation of scalefacs and thus a little
more noise shaping.
the quantization routines have been broken up a little:
quantize-pvt.c: utility routins like compute_ath and calc_xmin
quantize.c: single channel iterattion_loop and vbr_iteration_loop.
loopold.c: old dual channel iteration_loop (when -h is not used)
vbrquantize.c: new vbr code (not yet working) which will compute
scalefactors directly, and then figures out how many
bits are needed to encode the scalefactors.
(no calls to outer_loop!)
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
