Questions about resyncing audio and video

[JimmyBarnes]

When you have an AVI with AV out-of-sync, it is usually painfully obvious.

My attempts to fix this usually amount to remuxing audio and video in VDubMod and changing the Audio skew correction (Delay audio track by) in increments of 100 ms.

i.e. 0 for the original AVI, then say +100 ms, +200 ms, +300 ms and -100 ms, -200 ms, -300 ms

then visually comparing the AV sync in the 6 new AVIs.

I have 2 questions:
1. Is there a better way than this to fix out-of-sync?

2. If +100 ms is the "correct" adjustment, then in the AVIs with -100,-200,-300 ms it will be very obvious they are wrong. But when I compare +200 and +300 ms with +100 ms, it is hard to tell the difference, they ALL look correct.

Can anyone explain to me why this is, when even a difference of 100 ms should be detectable?

I have noticed this on a number of occasions and with several clips, so I think I am not imagining it.

[type1encryption]

with MPC HC and VLC there are options to adjust audio delay on the fly, you can in/decrease it in 10ms increments and look what's the right value for the specific clip

after finding out the right value you can remux the files without the necessity of muxing 6 different files

if you can't hear/see the difference between i.e -150 and -250ms (but -130 and -280 are out of sync) just use -200ms. It probably depends on experience with such files if you can't see the difference of 100ms delay, maybe you don't look at the right scenes (I am quite allergic to differences up to 50-80ms or so in close up dialogue scenes)

[CWR03]

I also use MPC to check delay. Right-click the video window, select Audio > Options..., check the box beside Audio time shift, enter a value and click Apply. When it looks right, you can enter the same value in VirtualDubMod.

Can anyone explain to me why this is, when even a difference of 100 ms should be detectable?
The hard part is finding something to use as a good synch point, such as someone slamming a hand on a table. Speech is a little harder, but if you can't tell between 100 and 200 it's probably just in between.

[JimmyBarnes]

after finding out the right value you can remux the files without the necessity of muxing 6 different files

if you can't hear/see the difference between i.e -150 and -250ms (but -130 and -280 are out of sync) just use -200ms. It probably depends on experience with such files if you can't see the difference of 100ms delay, maybe you don't look at the right scenes (I am quite allergic to differences up to 50-80ms or so in close up dialogue scenes)

The problem is, to use your own example, that there is no difference between -150 and -250, -130 is out-of-sync but -280 isn't. Maybe read the original post again carefully, as my numbers are rather different to yours but I have quoted yours in an effort to convey the problem I am facing.

In determining the problem as I originally outlined, I have looked quite carefully at sections throughout the whole where mouth movements are clearly visible.

[JimmyBarnes]

I also use MPC to check delay. Right-click the video window, select Audio > Options..., check the box beside Audio time shift, enter a value and click Apply. When it looks right, you can enter the same value in VirtualDubMod.


The hard part is finding something to use as a good synch point, such as someone slamming a hand on a table. Speech is a little harder, but if you can't tell between 100 and 200 it's probably just in between.

I understand, but surely by 300 it should be out-of-sync, but it isn't..

[type1encryption]

The problem is, to use your own example, that there is no difference between -150 and -250, -130 is out-of-sync but -280 isn't. Maybe read the original post again carefully, as my numbers are rather different to yours but I have quoted yours in an effort to convey the problem I am facing.

In determining the problem as I originally outlined, I have looked quite carefully at sections throughout the whole where mouth movements are clearly visible.

I'm just making the point, that 100ms _difference_ is not a problem for you, just use the median value, problem solved.