[Seth Bloombaum]

One idea – insert volume envelopes and lower her volume as soon as he starts to sing.

Perhaps you can shape her sound with EQ to lessen his voice in her mic.

The next is way more painful and maybe results wouldn’t be worth it, but if you’re feeling that you really need her trailing notes you can slice and dice any such notes that are “in the clear” and put them where you want. Hard to predict the results until you get into it.

If you really want to spend some time on this you could replicate his track, sync successive tracks wherever they need to sync to her mic according to where he is on the stage, and use envelopes to transition between his multiple tracks. This would probably be most successful phrase by phrase.

But check out how far the first two suggestions take you.

[Seth Bloombaum]

By a process of elimination that would seem to indicate that the audio was overmodulated at the transmitter (which is easy to do). (less likely – broken camcorder, head clog, error in transfer from camera tape to computer, impedance mismatch… peaks at -18db should be pretty safe)

I believe the G2 have an LCD display; when you set up a transmitter for a particular person you want them to speak as loudly as they might during a take, and adjust attenuation until their peaks just go to the (bad) meter reading. I’m not sure what that is on a G2, but if there is a mark about 2/3’s of the way up the scale you only want the occassional loudest peaks to briefly (instantaneously) reach above that.

Obviously it’s a safer setting to peak below the mark, but you need to do testing while listening to and recording the signal, monitor it very carefully to establish when it overmodulates, because too low a setting may tend to raise the overall noise. This testing should be done when you have access to good monitoring and some time on your hands (not on location!), it’s part of getting to know what your equipment will do.

Not sure about the G2, but on the previous generation of inexpensive Senn wireless the squelch control was pretty fussy too. Experiment with that and read the manual until you know what it does and when it does it.

Also, there is NO SUBSTITUTE for monitoring ALL the audio during a shoot. The overmodulation should have been caught at that time.