[FFmpeg-user] Some questions about PTS

[Paul B Mahol]

On 8/28/20, Michael Koch <astroelectronic at t-online.de> wrote:
> Am 28.08.2020 um 11:25 schrieb Edward Park:
>> Hello,
>>
>> I am not confident about this info but I've always thought the timebase is
>> usually the reciprocal of the framerate *or smaller*. As in, the duration
>> of a frame can be represented accurately enough as the difference between
>> the timestamps, which aren't counted using seconds, but "ticks" in
>> whatever timebase. So smaller fractions could be used as the timebase as
>> long as the system is capable, and on the other hand, when you create
>> low-fps video like 4fps, obviously the timebase isn't going to be 1/4,
>> it'll probably have the same timebase as any other export. (I think of it
>> as an analog wall clock, except it doesn't go tick-tock every second, it
>> goes tick-tock every 1/90000 seconds.)
>>
>> Actually, I think for the codec timebase, it is more common for it to be
>> 1/2 the reciprocal of the frame rate; if that's codec-specific, I don't
>> know why that is. Maybe you've also seen some setpts examples where you
>> divide/multiply something by 2 for some arcane reason? Hopefully someone
>> can explain further..
>>
>> When you delay some frames by whatever amount, it necessarily effects a
>> change in the frame rate (but not the timebase). I'm not sure where the FR
>> value for setpts comes from, maybe it wouldn't matter if it stays the same
>> as the nominal framerate if indicated by the media, but if it is something
>> that can change, maybe the effective rate at the end of the chain,
>> obviously it wouldn't work as expected.
>>
>> Just for the sake of curiosity, what has you looking to delay frames using
>> setpts? I feel there are easier methods.
>
> Let's assume the framerate is constant. For example, I want to delay a
> video by 5 frames and then hstack the original video and the delayed
> version:
>
> ffmpeg -i test.mp4 -vf
> "split[a][b];[b]setpts=PTS+5/(FR*TB)[c];[a][c]hstack" -y out.mp4
>
> Are there other / better / easier methods to do the same thing?
>

Yes, not using split filter, but using "-i test.mp4" twice.
Should generally use less memory for big offsets.

> Michael
>
> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-user
>
> To unsubscribe, visit link above, or email
> ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".