From time to time I watch additional material on DVDs I've bought chronicling the remastering of the film's video and audio. The audio remastering portion usually mentions something like "We were unable to find the dialogue masters for the film so we had to use the composite audio tracks for those sections of the film" but then in other portions mention "we had access to the music and sound effects tracks and original soundtrack masters". My question is simple: If they have a composite audio track with all the elements (music/effects/dialogue) and a separate music and effects track, isn't there a way to simply subtract the music and effects track from the track that includes the dialogue? I know some of you are going to say that there will always be a problem with phasing and/or problems with the fidelity of both tracks having to match completely, but I have a hard time that in this day and age to believe that there are no processes that can "adjust" two tracks to "match" each other. If enyone can enlighten me I'd sure appreciate it.

In a perfect world, where the 2 track stereo, or 5.1 mix was mixed from completely dry elements in the digital (or even analog) multitrack, then lining up the audio so it is sample accurate, in a Digital Audio Workstation, phase inverting either the dialog, or the rest will in theory result in dialog without the rest of the soundtrack.

In theory.

There are lots of reasons why this would not work in reality. If some ambience has been processed in at the audio mastering stage, this will smear elements of the dialog across the stereo soundstage, both in terms of stereo position and temporal position (in the case of reverb), so canceling out the dry vocal can leave a ghost image of the voice channel in the music/foley mix.

If the original audio was done in analog simple variation in tape player and recorder speeds (wow/flutter) can mean that a sample accurate version of the multitrack might not be possible.

Audio restoration is a field in its own right, and companies like Cedar specialize in it completely.

Our ears are capable of discriminating elements of sound psycho-acoustically (in that we can just 'ignore' some elements and 'enhance' others (In the same way as you can hear your own name across a crowded room - Whereas a computer algorithm would just interpret that as noise.

That's not to say that processing may not be able to do that in the future, just now, it's actually quite difficult, given the sometimes complex nature of the composite soundtrack.