Matching audio streams from two movies

[nji9nji9]

Thank you for the explanation.

However... I'm sorry... that is definitely wrong for the case I'm describing (see my very first post).

This is what your correction actually does:

[Attachment 89202 - Click to enlarge]

And when I wrote...

I muxed it, and checked the result... it seems to be pretty good

... I seem to have checked by chance at timestamps where it is close.

[ProWo]

Why should my approach in post #1 definitely be wrong for the purpose?
What are you trying to achieve with this graphic (post #31)?
Is my audio synchronized or not?
I muxed it myself with the mp4 in post#22 and it's in perfect sync.

[nji9nji9]

The red line in post#31 shows your correction you described in post#30.

I.e. your correction is perfect only at the intersection of the two graphs.

I rechecked the video's audio with your audio in detail at the timestamps
- with an audio editor this time.
There should be big differences at e.g. 17:50 and 48:03:
desync should be abt. 1s at that timetimes. See my graph.

However... the differences between the audios are only abt. 0.3s and 0.25s.
Which means that you didn't only stretch to equal lenght according
to end point peaks.
It seems that what you call "fine tuning" does most of the work,
i.e. lots of additional arbitrary corrections?

[ProWo]

The red line in post#31 shows your correction you described in post#30.
I.e. your correction is perfect only at the intersection of the two graphs.

I rechecked the video's audio with your audio in detail at the timestamps
- with an audio editor this time.
There should be big differences at e.g. 17:50 and 48:03:
desync should be abt. 1s at that timetimes. See my graph.

Have you watched the video besides your graphs?
Perfectly synchronized at 17:50 and 48:03.

[Apolosky]

The best thing to do in these situations, where manual sync adjustments are required, is, as previously suggested, to have both videos and review the frames. If both sources are properly synchronized, aligning the frames will give you the best result.

With the available footage, the best thing to do is what the master ProWo did: edit them in an editor where you can see the waveform or spectrum and adjust as you progress through the timeline.

As a tip, when you realize that the synchronized audio at the beginning and end doesn't match, you should consider two things. One is that it has a different speed (as mentioned with the frame rate), and the other is that there were some cuts in the silences or the silences don't last the same amount of time. But in this case, reviewing those two situations, neither was the cause of the offset.

One of the users correctly analyzed that the offset was caused by indiscriminate cuts in some scene changes. Analyzing them with waveforms is more complex. As I mentioned before, this is easier to do by comparing frames.

And finally, for me, with older audios, you have to take into account that synchronization wasn't as sophisticated back then. I've seen many audios with both transient and lip-sync issues. Current audio doesn't necessarily have native synchronization issues, but it's less common. Therefore, I don't think it's advisable to synchronize (in this case) with great precision; a 50ms or 100ms range works very well.

By the way, the audio provided by the ProWo master shows very good synchronization. Of course, you can take it and make some further adjustments to the spectrum, as you mentioned before, but you won't find anything automatic that does that.

Greetings!