[FFmpeg-devel] [PATCH] flvenc: prefer adpcm_swf after mp3

[Rich Felker]

On Wed, Nov 14, 2007 at 09:53:07PM +0100, Michael Niedermayer wrote:
> Hi
> 
> On Wed, Nov 14, 2007 at 07:20:16PM +0100, Alex Beregszaszi wrote:
> > On Wed, 2007-11-14 at 13:19 -0500, Rich Felker wrote:
> > > On Wed, Nov 14, 2007 at 08:40:12AM -0800, Mike Melanson wrote:
> > > > Alex Beregszaszi wrote:
> > > > > The only solutions is to have an mp3 encoder inside ffmpeg :)
> > > > 
> > > > One day, I'm certain.
> > > 
> > > Anyone up for implementing my idea: patching the mp2 encoder to
> > > support translating the output bitstream to mp3 syntax while not
> > > actually using any mp3 features?
> 
> what mp3 features?
> IIRC theres not much optional
> mp3 is just mp2 with MDCT +  antialias + different bitstream packing
> the MDCT we have, antialias is 2 lines of C code, just look at the
> antialias code in the decoder this is trivial
> 
> the only optional thing i remember is a silly buffer to store parts of
> the bitstream, we dont need to use that

Yes, bit reservoire. I would agree totally that a good encoder should
never use this misfeature (which makes all frames semi-P-frames). If
you want VBR, make a real VBR mistream. Otherwise live with true CBR.

Rich