I have some lectures on audio cassette I would like to convert to CD. The problem is that sometimes the lecturer's speaking volume varies greatly within a lecture. I'd like to get the volume levels consistent across the entire speech, with very little variation.

I accomplished this by applying Audacity's dynamic range compression, but there's a problem with that: the compressor reduces the volume of the loud parts (usually what one uses it for =). I reduced the loud parts quite a bit, but they were then all that much closer in volume to the background noise present in the auditorium. When I then amplified the entire speech, the background noise was amplified across the length of it.

I need a compressor that keeps the loud parts at the same volume and increases the volume of quiet parts to that same level. (I don't really care if some of the quieter parts, when boosted, have some audible background noise, because in general there are far fewer quiet parts in the lectures.)

I know I *could* look through the entire file and boost all the quiet segments one at a time, but this would get extremely tiresome since I've got quiet a few lectures I'd like to convert.

Any ideas?

If you can get yourself a copy of cool edit (it isn't free and I believe it is now called Adobe Audition), you can actually "draw" your own compression curve to suit your needs. In Cool Edit v1.2a, use the Transform->Amplitude->Dynamics Processing menu option. Sometimes you can really tweak the curve to actually perform expansion below a certain threshold to get rid of the background noise.

Hope this helps.