Re: [Openpvr-devel] filter chains, etc.

[Dave C. <de...@co...>]

Brian J. Murrell wrote:
> On Wed, Mar 27, 2002 at 09:14:24PM -0600, Dave Caplinger wrote:
> >>1) Capture video to lossless-compressed AVI (e.g. huffYVUY)
> What's the compression ratios like?

I don't know the exact numbers off the top of my head, but I believe 
about 2:1 is the best you can really hope for for lossless.

>>Filtering and re-encoding is likely to be time 
>>consuming,
> 
> I'll say!  I record real-time at about 40% CPU with mp1e and
> re-encoding with mencoder runs at 7-9fps, twice (2 pass encoding)
> which really amounts to 3.5-4.5 fps.

I'm thinking of this from a "regular end user" perspective, where they 
(i.e. my wife) aren't going to want to have to manually reencode stuff. 
:-)  But thinking this through for a moment: if this really was 
worthwhile, would it be useful to limit captures to 15 minutes per file? 
  The reason I say this is that if you're going to have to filter and 
reencode during "idle" times without any user intervention, you're going 
to need scratch disk space set aside for the intermediate files (which 
means that space can never be used to record into), so you want to 
minimize the amount of set-aside disk space, and also you want to try to 
minimize the time spent on the re-encoding because the user might come 
back (or a scheduled event might trigger) and need to record again, so 
you want to be able to free up the CPU to do that more important work.

So if you recorded in 15-minute (or even smaller, like 5-minute?) 
segments, and for any segment you were in the process of re-encoding you 
kept the original around until the re-encoding was done, the user can 
still play the segment that is being currently re-encoded (by suspending 
the re-encoding and playing the original), and then later the 
re-encoding can pick up where it left off.  Any single re-encoding task 
isn't going to run for tens of hours tying up tens of GB of disk space, 
but eventually the recoding (and thus disk space savings) will in fact 
happen.

This would all be predicated on the playback agent being able to peice 
together the segments (which could be completely different codecs) 
without any a/v-sync hiccups so the user would never notice it.  Seems 
like a big caveat to me, but I'm no expert at this.

Does this seem feasible?

- Dave