[FFmpeg-user] Image extraction broken with ffmpeg, but not with mencoder. How can I fix this?

Nicolas George nicolas.george at normalesup.org
Thu Sep 5 19:12:33 CEST 2013


Le nonidi 19 fructidor, an CCXXI, James Board a écrit :
> But I have a unique and unusual requirement.  I want to do frame-exact
> editing.

That is in no way unique or unusual. You are just doing it wrong.

> I split the file into a sequence of frames so I know at exactly which
> frame the cuts should begin and at which frame they should end.

You should not be doing that.

> Okay good.  All I need is a formula for determining what the timestamp
> is a for a specific frame (and vice versa).  What is that formula?

The formula is: timestamp. The timestamp is what identifies the frame, there
is nothing else, you should not be trying to compute it from anything else.

> >> 3: Timestamps are floating-point numbers, and floating-point numbers are
> >> not exact.
> >That is untrue.
> Really?  You might want to rewrite the Wikipedia page on floating-point
> numbers:
>     http://en.wikipedia.org/wiki/Floating-point_arithmetic
> Those guys think floating-point numbers aren't exact.

Do not try to play smart, you are losing. There was two parts in your
sentence.

> Bummer.   You had my hopes up earlier when you said each frame can
> be identified by it's timestamp.  So which is it?

The timestamp, no more no less.

> Either way, unless I can uniquely and consistently determine the frame
> number from the timestamp, and also determine the timestamp from the
> frame number, I don't see how I can do frame exact editing with time stamps.
> That's why I use frame numbers.

Well, I will not waste any more time on you. Good luck.

Regards,

-- 
  Nicolas George
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-user/attachments/20130905/a3d41519/attachment.asc>


More information about the ffmpeg-user mailing list