Re: Chapter marks not where specified.

On Fri, Mar 30, 2012 at 4:30 PM, Tom Brock <tomb-yZYHXxUkMT10YYv+nEap2lpr/1R2p/[email protected]> wrote:
> I think you mean one gop per frame

?  GOP = Group Of Pictures -- not sure how you get one gop/frame
unless you have all I-frames.

> read, that is not a requirement: it is simply typically the case.  A VOBU must
> contain between 0.4 and 1 second of playback.  In one case, I've seen ffmpeg
> produce 5 I-frames in 0.32 second, and there's no indication that it thereby
> violates the DVD standard.

Tough to say without the DVD spec ($$$).  I initially wrote dvdauthor
for two reasons: one, because I prefer Linux, and two, because I
wanted to import videos (without reencoding) that weren't 100% correct
-- I had 544x480 NTSC mpeg-2 videos off a TiVo.  One player would deal
with it properly, another wouldn't.  Then I tried piecing together
chapters from separate video clips; one dvd player would play
correctly, another would have a/v sync issues over time (since the
timespan of an mpeg layer-2 audio frame is different than the timespan
of an NTSC video frame).  Eventually I decided you were pushing your
luck if you violated "normal" conventions (I had to guess what normal
meant, since I wasn't willing to pay for the spec); hence the set of
guidelines I laid out.

FFMpeg has always been problematic when it comes to producing streams
that work well on hardware players.  I strongly suggest you use
mjpegtools instead.  I think I sent a patch to mjpegtools in order to
support chapter boundaries without violating GOP lengths (i.e. if a
chapter was coming up in 21 frames, and the GOP length was set to
9-18, then it would do a GOP of 12 followed by a GOP of 9, rather than
a GOP of 18 followed by a GOP of 3).  I've never seen mjpegtools
insert a VOBU at any location other than at an I-frame; I stopped
tracking what ffmpeg does a long time ago -- ran into too many issues
with both the encoder and multiplexer.

>> Other "shoulds" include 2 B frames per P frame, minimum GOP length of
>> (I think) 9 frames, and max of (I think) 18 frames -- though that
>> might be NTSC dependent.
>
> It may be that dvdauthor cannot place the markers as
> specified if they do not align with VOBU boundaries in the input file

I think that is true -- I think that's simply the way the navigation
fields work in the VOBU.  It is questionable whether your hardware
player will be unhappy if your VOBU boundary does not correspond to an
I frame, and at this point I don't remember what assumptions dvdauthor
makes about VOBU boundaries starting on I frames.

> On the other hand it may be that dvdauthor is
> intended to be capable of adjusting VOBU boundaries when necessary, so as to
> align with chapter markers, but fails, for reasons at present unknown, to do
> so, in which case it should be fixed.

As you've noticed, it tries to round to the nearest VOBU.  I'm
surprised if you've seen it be substantially off, ASSUMING you have
VOBU's at I-frame boundaries.
-- 
        Scott

------------------------------------------------------------------------------
This SF email is sponsosred by:
Try Windows Azure free for 90 days Click Here 
http://p.sf.net/sfu/sfd2d-msazure