[mythtv] Lossless MPEG2 editing!

[Cory Papenfuss]

> Sorry to chime in... but I posted a bug report to point out that 
> mythtranscode does not work for me with cutlists (it does a great job of 
> transcoding RTJPEG->MPEG4 though). Is there another way to do that (almost as 
> easy as transcoding directly from the frontend)?
> Thanks,
> Manu

 	The quick and most accurate answer is probably "No."  I'm not 
completely sure what you're asking for.  A program for hand-editing out 
commercials is avidemux... works pretty well.  It doesn't need to 
reencode, but it can.

 	Regarding lossless MPEG2 cutting, there are a number of issues:

- Funky MPEG2 streams... between ivtv/DVB/HDTV/Dig Cable, there are lots 
of different bits within a "standard mpeg2 stream."  There can be *lots* 
of other streams and formats of wrappers, packets, substreams, etc all 
under the umbrella of an "MPEG2 capture"

- GOP vs. non-GOP accurate edits.  MPEG2 streams have I, P, and B frames. 
Only the I frames can "stand alone."  The other two rely on 
previous/future combinations of other frames to generate a frame.  That 
means that cutting arbitrary frames is difficult to make a lossless 
operation.  Some re-encoding around the cutpoints is usually necessary, 
even if cutting on GOP boudaries.  Read the GOPchop docs for more on this.

- GOP length:  Real-time encoders (e.g. ivtv-based cards) don't have the 
benefit of noncausal filtering (read: they can't predict the future), so 
they make compromises quality of the video for a given bitrate.  AFAIK, 
the hauppauge cards and ivtv driver combination only produce one GOP 
sequence, which is about 15 frames long.  That's roughly 1/2 second.

- Sync: Audio and video packets don't have to be very close to each other. 
It's probably likely that a good MPEG2 stream cutter will also have to be 
a rudimentary player to figure out which streams and packets are really 
necessary to fully decode something.  Since MPEG2 streams are both 
pseudo-realtime and packetized, they have Presentation Time Stamps 
embedded to allow for prebuffering of data.  This can then be "presented" 
at the correct time.  If you take a chunk out of the middle, all PTSs past 
this point now need to be tweaked.

 	I'm sure I'm forgetting some things, but those are some that I've 
run across.  If it were easy, it'd be done by now.... :)

-Cory


*************************************************************************
* Cory Papenfuss							*
* Electrical Engineering candidate Ph.D. graduate student               *
* Virginia Polytechnic Institute and State University 			*
*************************************************************************