[MEncoder-users] two-pass encoding for creating DVD-video

[Nico Sabbi]

Trent Piepho wrote:

> It doesn't look like this fix is integrated into the 2-pass rate control
> system.  My understanding of that patch, is that if a frame is encoded and the
> resulting size of that frame breaks the maxrate (by overflowing the rc
> buffer), that one frame is re-encoded at lower quality until it's small enough
> to not overflow.  It's much better to instead drop the quality of the frames
> before by a little bit, rather than drop the quality of one frame by a lot.
> That's sort of the whole point of 2-pass rate control.

if so why don't you report it to ffmpeg-devel?

> 
> Lots of people would claim this bug didn't exist for a number of reasons:
> 
> Exceeding the maxrate usually has no effect for playback on a computer.  It's
> only hardware DVD or SVCD players that care.  Someone would encode to mpeg2
> and the final file size would be correct and it would play back fine with
> mplayer, so they assume there must be no problem.  It's easy to jump to the
> conclusion of user error.
> 
> The maxrate allowed for DVDs is quite high and many hardware DVD players can
> do at least a little more than the max allowed in the DVD specs.  It's not
> that hard to encode something (especially from a high quality source, like a
> ripped DVD, as opposed to an analog tv capture) and not exceed the maxrate.
> It is a much bigger problem for SVCD, because of the much lower max bitrate
> allowed from CD-R media vs DVD-R media.  It's easy for someone to encode a DVD
> and have it work in their player even when vrc_maxrate doesn't work.

overflows are very easy to spot during encoding: the muxer will almost 
unavoidably complain of a dts underrun and raise the muxrate;
in order to realize it users should at least pay attention to the 
report, but most of the times they won't notice it because they are 
110% clueless