The first third of the MPEG-1 file I have is in synch. The last third is in synch. The middle third is not. I can imagine it's relatively easy to shift the audio of the whole file but how can I mark a portion of audio and shift just that much?
No, it's not easy at all.
You must the sound as uncompressed WAV, as in 3) of http://forums.afterdawn.com/thread_view.cfm/129217 . Probably you'll have to re-encode that mpg in another 'dummy' movie [= you can put 'motion search precision = lowest quality (very fast)' , since you won't keep the video] with TMPGenc, which is needed only to have a 'standard WAV' instead of a corrupted one, if EAC cannot load the 1st WAV you extracted with VirtualDub from the input mpeg, modify cutting(inserting silence and/or cutting that WAV with Exact Audio Copy - EAC (Tools__Process WAV) moditying it 'manuallly'.
After that, save the WAV , compresss it to 'mp2 for SVCD' , be sure it remains 44.1 Hz, and multiplex it to the video.
If you dont' manage fo compress it in MP2 44.1 kHz (HeadAC3he or BeSweet+GUI), make with TMPGenc another 'dummy' file by using that modified WAV as 'audio source' (video source = 'video input') and multiplex the input movie (video) to the 'dummy file' (audio) to create the final movie.
But it's very hard, you'll have to make many test with 'dummy movies' until you manage to hae a modified WAV which is exactly in-sync with the whole movie.
Ah, to save time you might split the mpeg (TMPGenc File__MPEG Tools__Merge & Cut or, if it cannot do it in the corrupted part, VirtualDub --> Video__Source Range but you'll need to compress the video to AVI, first, keeping the in-sync part separated by the non in-sync part, them doing that difficult process only in the non in-sync part. You can even add a delay to the MP2 with the application I suggested, so you can avoid the 'edit the WAV' part. But probably the WAV non-in-sync part can be loaded only if you make the 'dummy' trick to have a standard WAV file.
Good luck...
No, the method works very well.
The only thing to do is doing carefully some check of the 'edited WAV' , keeping 'motion search precision = lowest quality (very fast)' with TMPGenc. If you're sure the sound shift is constant, only a piece of the movie is enough.
And separating the the 'in-sync' piece of movie movie from 'out-of-sync' part of it is another trick to speed things...later you can re-join them quickly.
