View Single Post
  #8  
Old October 10th 18, 10:37 AM posted to alt.windows7.general
NY
external usenet poster
 
Posts: 586
Default question for the video editor folks

"J. P. Gilliver (John)" wrote in message
...
I've looked at the Wikipedia article and it gives the impression that the
GOP size is generally fixed; that surprised me - I'd always assumed they
varied with content (e. g. a scene cut requiring a new I frame, a static
picture not needing any after the first).


Yes, my understanding is that GOPs are normally a fixed spacing, especially
on DVDs, but they are a *maximum* number of intermediate differential frames
before the next full frame, and that it makes a lot of sense to put a new
full frame at every shot change, to avoid having huuuuge difference frames
between the last full one and the current frame of a new shot. Broadcast TV
seems to be more tolerate of larger GOPs than DVDs which impose a smaller
maximum GOP size: when I've been encoding recordings of home movies to put
on DVD, I've seen a message "GOP size too high - will need partial recoding"
in VideoRedo.

As I understand it, you can cut on any frame boundary, even joining two
unrelated differential frames, as long as your software generates new full
frames (by taking the last full one in the second clip and applying all the
subsequent differences up to the "out" cut point) and then calculates brand
new difference frames from that point to the next full frame.

AIUI that's how Video Redo makes seamless edits which don't have to be on
GOP boundaries. You can even see on the status dialogue, it says "copying
frames" and then briefly, at each edit point "recoding frames". I presume it
is this re-encoding around the edit point that Avidemux and other
FFMPEG-based software doesn't do properly.

VRD even has a setting which allows you to vary the bit rate of the
generated frames. I noticed when I was removing the commercials from
low-bitrate, low-resolution (544x576) recordings on Yesterday, Drama etc, I
sometimes got a lot of compression noise at the join. I was advised by VRD
support to set a per-recording value of minimum bitrate which is ignored for
the majority of the recording when you are just copying source to
destination, but is used when any full and difference frames have to be
generated.

(Note that I use terms "full" and "difference" because I can never remember
which of those are termed I and which are termed P or B! I'd thought it was
P for picture (full) and I for intermediate, but from comments up-thread I
think I may be wrong.)

Ads