[mythtv-users] choosing an archive format for recordings

[Cory Papenfuss]

 	CPU usage is a bit on the high side.  I've got two machines at school 
that I use, and each is a dual processor.  Mine's an athlon 2400 (2.0 GHz), and 
the other is a P4 2GHz.  They both take roughly 4 hours to process a 1 hour 
show (~43 minutes after commercials cut) with the 2-pass, resize, and denoise. 
Avidemux only uses 1 cpu per instance, so I can run 4 at once with minimal 
slowdown... thus the script.

 	I have not looked at the scripting capabilities.  I still wonder if 
avidemux is the right direction for a fully mythtv-integrated, backend-like 
transcoding setup.  It's great for hand-cutting because it's got the GUI and a 
bunch of easily-tweakable options.  For special-purpose stuff, I think the 
command-line 'transcode' package is probably more powerful and robust.

 	WRT the I-frames, I usually cut on an I-frame closest to the commercial 
entry/exit.  In the process I've noticed that the PVR-250 with ivtv is not too 
smart when it comes to GOP choices.  I'm sure it would take a lot more 
buffering and decision-making on the part of the card to determine the best 
time to send an I-frame (e.g. a scene cut).  I've also seen an I-frame split 
across fields of high-motion interlaced scene.  That's pretty much why I encode 
at higher bitrates and then downconvert later for space... more time to chew on 
stuff offline so it can be packed tighter better.  Of course, since we're 
reencoding offline anyway, cutting on a I-frame shouldn't be necessary.

 	As far as I know, MPEG is strictly streaming.  There's no notion of a 
"completed chunk."  That's one of the reasons for avidemuxing having to index 
the thing before he works on it.  He's gotta know where everything is, and how 
big to make his sliders... :)

-Cory

*************************************************************************
* Cory Papenfuss							*
* Electrical Engineering Ph.D. Graduate Student                         *
* Virginia Polytechnic Institute and State University 			*
*************************************************************************


On Fri, 6 Aug 2004, Leo Weppelman wrote:

> On Thu, Aug 05, 2004 at 10:22:34AM -0400, Cory Papenfuss wrote:
>> On Thu, 5 Aug 2004, Leo Weppelman wrote:
>
>> 	I'm already using the command-line version of avidemux to do my
>> processing.  I pretty much just use the GUI to find the commercial
>> cutpoints. The config files and cutpoint documents are pretty
>> straight-forward, and I sorta hacked up a script (attached) to actually
>> process the files.  It goes through a list of ".edl" files which contain
>> the cutpoints, etc saved via avidemux and has contents similar to:
>
> Thanks for the script! I have been playing with avidemux and your script
> a little bit. The bottleneck is the cpu :-(,  so the amount of 'playing' is
> limited. But instead of generating '.edl' files, it might be better to use
> the scripting capabilities. You can set the markers, but you cannot delete.
> But that shouldn't be too difficult to implement... Another thing is that
> The myth cutlist is not (neccessarely) on I-frame boundaries. The doc's
> suggest that it is much better to make sure it is (although I have no
> idea how bad it looks if we don't) . So that is a point to keep in mind
> too... Positive thing is that Myth and avidemux seem to be agreeing on
> the frame numbers. I'll see if I can manage to build avidemux from source
> today...
>
> Do you know how the running time of the mpeg can be determined? We need
> to know the running time after cutting. That way the wanted size after
> the encoding can be determined.
>
> Leo.
>