I'm far from an expert, but I'll take a shot at some questions.
I'm not a DIy'er and the examples and instructions are daunting.
Given the pre-recorded samples on the ambio site, sampling the effect isn't really that tough. Simply temporarily move a set of speakers out front with about 12" between them, and listen at about a 6' distance. Not perfect, but it'll give you a taste.
In my case (see gallery or system) I have a 10 ft hidef video screen that can't be obscured by two closely spaced left-center and right-center speakers.
Yeah, this would be a pretty insurmountable hurdle. Ambio and video don't really mix well as far as I can tell - ambio is pretty much a single-person sweet-spot approach.
* Is image specificity/density good on Ambiophonics? What I mean is, when listening to recordings that push the enveloped with crosstalk cancellation, such as Q-sound recordings like Roger Waters' Amused To Death, the soundstage is fabulously wide at times, but the effect is very unnatural and quite a bit phasey, even when done well. The dog barking off axis is cool, but not exactly specific to a point in space. I'd hate to lose solid imaging to gain 120 degree soundstaging....
This will be highly recording dependent, but in my experiments good recordings maintained solidity and density, but the impression of being in a listening room just vanished. For relatively natural recordings where the content is fairly center-weighted, things worked well (e.g. Cowboy Junkies Trinity Sessions). For more artificial recordings, solidity did suffer sometimes - for example Rodrigo Y Gabriella 'Tamacun' became cartoonish because one guitarist was way-over-there and the other was way-over-the-other-way, and spatially integrating these across a 120-150 degree arc just didn't work. (this was disappointing, as Tamacun sound spectacularly good on my 'conventional' set up)
Remember that Q-Sound has a big barrier to overcome - trying to to xtalk cancellation in a generic/arbitrary way with a presumed 60-degree separation between speakers. Getting it to work at all is pretty impressive.
* do the fronts have to be dipoles? I'd have a chance to play around with this in my newly cobbled-together home office setup (pc readily available, etc.), but the speakers are the nice but direct radiating VMPS 626R's (edit: did research and found out that the contrary, that they could be as point-source as possible).
Your edit is correct - I think there is a confusion of terms here. In ambio the 'stereo dipole' refers to the physical arrangement of the two front speakers, not the design of the speakers themselves. I think having actual dipole fronts would be somewhat problematic in anything but very large rooms. (Another source of confusion may be that line source planar/ribbon speakers work well due to small physical extent, and these frequently are dipole in nature)
* could the post-processed wav music be eventually delivered via a Squeezebox/streamer, or would this require a nasty set of Squeeze Center plug ins?
In theory, no reason this couldn't work. The samples on the ambio site are done this way - pre-processed so that they can be played back without processing. It may take some work to set up a batch-convert process, but for example the RACE implementation based on AudioMulch would probably work fine that way. You would want to ensure a 24-bit output file format due to the processing/computation, so your output filter and playback chain would need to support that.
