I just created a couple of mock-ups of what an interface for the audio diff feature proposed for GSoC might look like. I am posting this to the mailing list in addition to my GSoC proposal. Feedback will be appreciated!
Here's a quick mockup of one idea about how the diff interface might look like: http://mercurysquad.googlepages.com/audacity_mockup_expanded.png
And in the condensed form: http://mercurysquad.googlepages.com/audacity_mockup_collapsed.png
The expanded image has labels describing each element of the UI. There can be a new "Diff regions" track to facilitate operation. In the first stage, I expect the diff algorithm to operate on pairs of tracks, identifying regions which are the same, shifted or inserted/deleted. (The shifted part is a post-processing step after the diff has found the indels). Each of those regions will be marked on the diff track, and clicking each marker will select that region of audio as a regular Audacity selection which can be copied, cut, effects applied etc. The corresponding region on the other track is also shown with, e.g. red highlighting to give a visual indicator of where on the other track it can be found. I expect that multiple such regions could be selected, turning them into regular audacity selections which can be manipulated easily. An additional idea is to divide each track into clips based on the diff regions, though it might introduce gaps in the audio (?).
The basic use-case scenario would be an audio engineer trying to create the best audio mix from several different 'takes'. This is the scenario I (and perhaps most others) find most useful and straight-forward to implement. Such an engineer can then clearly see how two clips differ, which parts, and in which ways; to finally choose the ones which sound the best (subjective) and/or are the least/most different (objective).