wonder if using something like acoustic mirror would help?
Spot?

Don't waste your time. To accomplish this task with any kind of acceptable results is simply impossible.

John

Time to call the lip reader. Long time looping....

Try using a parametric equalizer to tune it - often 3 or 4 sections - and sharp rolloff below 90 Hz or so to get rid of room nodes.

Then gentle downward expansion, with one band for formants (under 900 Hz), one for noises (next octave), one for consonants (1.7 kHz - 4 kHz), and one for upper harmonics.

Next use a sampled noise reduction tool if the above doesn't work well enough. Don't use NR before the above.

John

Very high end noise reduction tools can get a lot of the way there, but any remaining dialog usually artifacts pretty badly. Sounds like it's under water in many cases. I do a lot of this for law enforcement and attorneys, where the only objective is to make dialog intelligible enough for transcription, as opposed to having it be suitable for broadcast or exhibition, for example. Having said that, there ARE tools available to the forensics community that are a lot better at tasks like removing reverb. I suppose you could go to someone certified for heavy forensics work to help with the problem.

I would think a noise gate would help (especially if you could add a music bed).
Randy

Not with echo.

John

My theory upon writing that post (when I was still half asleep) was that if you set the threshold to cut off the audio right after the person spoke then one wouldn't hear the following echoes, but now that I think about it I guess the problem is the echoes lapping over the audio....derrr... therefore, it appears I'm full of crapola....never mind : )
Randy