Converting 25fps PAL to 24fps

[kropotkin]

We all now what inverse telecine does, however, i assume that it must be some kind of tehnique of converting 25fps PAL to 24fps, without losing the audio sync. Anyone did this with standard software?
I search the net and got little results. The only software which is supposed to do this, and poped out in my search, was the Cinema... whatever, the Mac based video editing software.

[DrP]

All I do is either edit the AVI header directly and change the frame and audio rates there using say ZTreeWin, or open the AVI with virtualdub, export the video with a frame rate change, export the audio to a .wav and use Audacity http://audacity.sourceforge.net to change the audio rate to match the video again. Bit of math is all thats required.

Now you might say 'Eh? Won't that cause a frequency shift in the audio?' .... sure will. But here's the nice part, it will usually result in the audio being set to the correct pitch! I was supprised to hear the difference in the audio pitch from a PAL version of a TV show to a NTSC version, where PAL was simply sped up to 25FPS and NTSC was done with a proper field manglation.

[kropotkin]

Ok i'll try, i have downloded audacity and i'll see what i can do tonight.

[kropotkin]

DrP, if you notice this post, maybe you can help me. I did what you suggested and i reencoded the movie at 24 fps, with no sound. The sound from the 25 fps movie was transformed with cool edit pro /time stretch, to the aproximate length of the new movie, in fact i substracted a frame every second and calculated the new length. However, the computation wasn't so precise, despite the fact that i use 3 digits in precision, and, when i reexported the movie with the new sound track, virtualdub said, at the fram rate, "change so video and audio match" sometihng like 24.0xx fps,. Even so, the movies wasn't syncronised. I also tried to force 24 fps, but the problem subsided.

[DrP]

The multiplier for the audio rate when going from 25 to 24 is 0.96. So 44000 becomes 42240. So all you should have to do is edit the wav file headers and insert that value for the sample rate or tell your audio editor that this is the sample rate. I know audacity lets you set any sample rate like this.

VirtualDub should then be able to make the composite AVI without problems. If it does complain about incorrect length vs the video, maybe try a 3rd party mux tool like avimux or even graphedit.

Normally the forums send me an email when someone replies to a thread I'm in.. it didn't for some reason, hence the long time to reply.

[kropotkin]

It is okay now, without your advice i managed to do the job by using the "resample" time shifting method in cool edit pro, where the editor modified both pitch and time. Thank you anyway, for your time. Now i have a bit different problem, which is somehow (i don't know) related to this one: two avis which were ment to be joined with this one have some problem. The strange thing is that, even the files are identical in video and sound settings, virtualdub reports that are different when i try to join them together. How can this be possible, when the same virtualdub said at "file information" that are the same?

[DrP]

You can get this problem if the video codec supports different internal compression types. huffyuv tags RGB compression differently to yuv compression (fair enuff, they are different). To you and me it says huffyuv, but Virtualdub won't let you append them.

Try joining the two video parts without the audio streams incase there is something odd going on.

If that doesn't work, the only way may well be to recompress both avi with say huffyuv (lossless) and then join the resulting files.

[kropotkin]

Okay i'll try your idea, with the no sound join, but i have a feeling that the last option (recompress) it is the only way. I figured that out myself, but now i'm really in a shortage of disk space, because of a bunch of recording a had to edit and i was forced to keep on the hdd.
However, the source of the material was recorded with avi_io picvideo mjpeg, and in fact are two segments created by this software. This is odd .. very odd.