Formula for NTSC to PAL lip-sync

[macnab]

Have AVI files in NTSC. Converting to VCD PAL.
I have experimented with one file and have the following info:

AVI (NTSC) is 51535 frames at 29.97fps = 1719.568sec.
Strip video and convert to PAL: 42988 frames at 25fps = 1719.520sec.
Strip audio: 1719.562sec.

By experimenting if I use time-stretch in my audio-editor to shorten the file by 420ms (a ratio of 0.000244) I get almost perfect lip-sync when I combine the video and audio into a VCD file.

I can't find any combination of the above figures which gives me a ratio close to 0.000244 (which is not the EXACT ratio I should be using.). Does someone know a formula I could use with the above figures to get the correct ratio for lip-sync. Experimenting is a lengthy process.

[Alex_ander]

The difference in length 0.048s for old/new video is very small (about one PAL frame duration), almost negligible for sync measurements by eye. If you felt the audio got out of sync with new video, then most likely, original audio had had a constant delay. It's absolutely unnecessary for synch to have video/audio lengths matched. So unlikely your compensation numbers can be derived from changed video length.

[NDMMackay]

Yes, and the difference in length of the extracted audio is just as small. And the source video is correct.

Maybe I should try 1) combining NTSC video with extracted audio to NTSC VCD; 2) combining NTSC video with extracted audio to NTSC MPEG-2; 3) combining PAL video with extracted audio to PAL MPEG-2; and see where the problem lies.

[Calidore]

Dividing 1719.52 by 1719.562 gets <>0.9999755752. The difference from 1 is 0.0000244248 Did your confusion come from accidentally dropping a leading 0 from the results?

Using your AVI length of 1719.568, we get <>0.999972086, or a difference of 0.000027914. Try that and see if it fixes the drift.


Best,

Calidore

[NDMMackay]

Using 1719.562 - 1719.52 you get 0.042 seconds difference in file length. I need a difference of 0.42, which is 10 times larger.
If you do it the division way, you get a ratio of 0.0000244248. I need a ratio of 0.000244, which is 10 times larger again (of course).
I fail to see where the factor of 10 comes in.

The difference between original video length and extracted audio length is 0.011 seconds, which is 0.25 of a frame.

BTW I did try recombining the NTSC video with original extracted audio to NTSC MPEG-2 file, and the lip-sync problem is there, so the error comes in when extracting the audio. It gives a file of the same length as the video, but somewhere there is a bit-rate problem, I have to time-squash it by a factor of 1.0024 to get lip-sync. And so in the final file, the audio runs out 0.42 seconds before the video does.