Well, dead silence isn't good (unless it's suspoed to be there), but normally people expect certain sounds in cerain situations. If you show it to someone & they don't say "what was that noise?" then you're allright. :)

Thanks guys. I'm going through the timeline again now and dropping the reef noise down to -inf when we have longish stretches of music. Previously it definitely would NOT have made someone say "what was that noise" but it just kind of rumbled a a bit irritatingly beneath the music, not really adding any value. I've got it surfacing quite high again during periods without voice or music, giving us that underwater feel. There are no short cuts to mixing audio properly. I've got literally thousands of keyframes in this project now, and climbing. Next step will be writing my own "clever ducker" script to ease the pain for future projects like this.

Ultimate S already has the "clever ducker" script if you have Ultimate S...
Keep in mind, you can run the music in stereo, using a spreader plugin, then run your bubbles/underwater sound in mono, panned slightly right or left, or using a "narrower" plugin to place it exactly, something like Q Sound.
Run your dialog somewhere in the neighborhood of 15dB louder than the water and the music. You can also keep the water sounds out, and just use it to fill in dead music spots, or between musical sections.

... and...
to me the whole video project is like an orchestrated dance.

I always tell my editors, "don't think - just try it out". I like a lot of variety in my film/video work. You are doing everything in your power to keep the audience awake and interested. So (and without seeing the exact video in question, this is hard) I'd definitely be in favor of a continual change of audio environment.
Don't forget that the majority of your end viewers will (and this is rapidly changing) be listening through a 3" to 5" speaker, generally placed way too close together to really "get" a stereo image.
Any pro working on either music or theatrical video,will always do two different mixes (assuming the budget is there) - one for the huge Voice ot the Theater speakers - the other for the automobile or home TV.

One (the big one with fine audio gear and enormous speaker set-ups) will have magnificent layering of sounds, music, SFX, on camera and voice over tracks, sensitively mixed to a fine puree of audio delight.

The second will basically be, "Can we understand what that guy is saying????"

Well, I overstate the situation above, but not as much as I'd like to believe.
With respect to your exact question - Definitely Lose the ambiance (especially if it's just an underwater bubble track" - OK - we Get It - there's a SCUBA tank around. Play with the music - Dance with it - do some synced cuts of your video to the track - THEN bring back the bubble track - maybe totally on it's own for the (for example) scary scene where your "star" may be entering some kind of trouble - then fade up the music, etc.

And "play" with the music itself! If you start with a Baroque lute and flute track, try a Salsa band next for contrast - "Ewehhhee -that's a horrible idea"

Whoa! You're thinking again! Wrong - try it out. Otherwise you are limiting yourself to what you know - and - (and this is for EVERYONE) - THAT AIN'T MUCH. Point and counterpoint - rythmn and cadence - surprise me. Make me laugh - shock me - delight me - ENTERTAIN ME! Trust me - that's what the best talents do. They try out crazy stuff, looking for wonderment.

If you have fun - there's a good chance your audience will have fun. If your video all looks the same - and likewise, if your audio all sounds the same - well - BORING.... technically correct but - dull.

Play, play - play - that's why we DO this stuff.
best,
v.

Fantastic advice guys. Thank you very much indeed.

I'm not sure I'm brave enough with this, my first project beyond personalised scuba DVD's and my first project with a voiceover to experiment too much. I just need to get a solid project out. Maybe I'll do a version 1.1 when I've got a bit more confidence.

Spot, I tried the Ultimate S ducker and I didn't like the way the duck was inboard of the voiceover event. I wanted it outboard (as per Excalibur). Actually even better would be for it to sense when the voiceover audio level reaches a certain threshold (user-settable) and make the bottom of the duck use that point because my voiceover clips have varying tops and tails outside the actual speaking.

Maybe there's a way for me to auto trim a batch of voiceover clips in Vegas or Soundforge???

Vicmilt has some ecellent advice there. I once had to put some music to a pig hunting scene. Now all I have that I own the rights to is a rather limited collection from SonicFire, so I tried some funk/rap. Well, what a hoot. Poor porker gets shot right when there's a "UhOoh" in th track and a temp change. The rest of the mob of pigs trot off in time to the music, magic stuff.
So yes, give the seemingly inappropriate a try, you never know how it'll turn out.
Bob.

For audio newbies like me, I think Spot was referring to QSound's mono-to-3D localizer, QX, part of their QTools/AX product. 80 bucks or so.

I guess, Spot, you meant spreading the stereo to make it more distinguishable from the mono rumble and voice. I couldn't find a spreader plugin within Vegas but do you think running Sound Forge's "Expand Stereo Image" preset(in Pan/Expand) would be a good way to go? I'm a bit worried about ruining my composers' lovely music!