> From: Takehiro Tominaga <[EMAIL PROTECTED]>
> Content-Type: Text/Plain; charset=us-ascii
> Date: Sat, 08 Jan 2000 23:15:26 +0900
>
> Inspired from CVS version of vbrquantize.c, I implemented the new VBR
> routine into my snapshot which is available at
>
> http://www.isoternet.org/~tominaga/lame-beta/lame-0108.tar.bz2
>
>
> It works and is faster than CBR(!), but still has to do some tuning
> (please don't encode the SQAM test file like gspi35_1.wav :p).
>
Great news! although finishing vbrquantize.c was one project I
was really looking forward to working on :-(.
I thought it would be faster than CBR, since all the
quantiziaion and lossless coding only has to be done once at
the very end.
Another thing I am interested in is trying this algorithm for
CBR. The simplest choice would be to fix all the scalefactors and
then very the global stepsize until the right bitrate was obtained,
but because of the nonlinear quantization effects, this is not
quite correct. To be most accurate, changing the global gain
would require recomputing the scalefactors.
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
