I couldn't resist fiddling with the sound and it made a big improvement. This was all done in Adobe Audition, but I am getting better with Soundtrack Pro and may do the next project with that.
Here are the steps:
- Noise reduction: Fortunately there was a quiet passage to use for the noise profile. That was a bit lucky though. Note to videographers: always record more than you think you need and make sure you get some "room tone" (background sound when no one is speaking). Noise reduction can be tricky but in this case it worked out well without a lot of fiddling.
- EQ: The voices sounded very thin and distant. I boosted a region around 200Hz by about 6dB to fatten them up a bit.
- Limiting: There were lots of door slams, microphone bumps, etc. The Levelator (see step 5) can cope with this, but some of the spikes were pretty extreme so I helped it out by doing a limiting step. I find that if I use the "gather statistics" option in Audition to limit no more than .05% of the samples, I can boost the gain a fair amount and it is inaudible.
- Bandpass filter: A technique I picked up from working with the Conversations Network is to routinely bandpass filter spoken word content between 80Hz-10KHz. There's no useful information outside that band and it just cleans things up a bit for downstream processing.
- Levelator: Last step was to apply The Levelator. It did a good job of evening out the quite varied levels of the presenters and making the audience Q&A audible.