[mythtv-users] Announcement: Looking for Alpha testers for a generic Lossless cut solution

[John P Poet]

On Thu, Oct 11, 2012 at 11:58 AM, Doug Vaughan <r.d.vaughan at rogers.com> wrote:
> Dan,
>     I just checked the showinfo filter out and you are right I cannot find a
> reliable correlation between what ffmpeg indicates as a keyframe and what
> MythTV put in the seektable when recording. Ffmpeg indicates there are
> significantly more keyframes than MythTV identified. The ffmpeg keyframe
> numbers are not close to what is in the seektable. I tried to see if there
> was some sort of base shift number that could be used in a formula but
> nothing was apparent.
>
>     For my HD-PVR 720p recordings the keyframes are almost always 128 frames
> apart starting after the first non-zero keyframe.  I noticed that ffmpeg
> shows a consistent 32 frames apart. Even though 32 is a multiple of 128 it
> still looks too inconsistent to extrapolate. I think the 32 frame number is
> the same as what mythcommflag --rebuild inserted for the same HD-PVR 720p
> recordings. I would have to check but am too busy right now.
>
> Interesting but does not help.

Been a couple of years since I looked at this, but my memory is that
the HD-PVR generates an "I frame" every 32 frames, but only generates
and IDR every 128 frames.  I think with mpeg you can get away with
using an "I frame" as a key frame, but with H.264 you can only use IDR
and SEI frames as key frames.

I believe that ffmpeg is counting I frames as key frames, but if you
try to start playback on a HD-PVR I frame, you wont get a "complete"
picture.   A "complete" picture can be built up from an "I frame" plus
(something), but I don't think ffmpeg does that correctly with the
HD-PVR H.264 stream.

Anyway, that is all from my foggy memory so may not be completely accurate...


John
-- 
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?