I shooting a small crowd scene of anywhere from 10-20 people. I need to somehow use the natural recorded audio and ambience but make most of the recorded voices somehow unintelligible, sort of like walla, but it still has to sound natural.
Without listening to your audio or knowing how exactly it was recorded, I would try the following:
1. Sweep a wide Q filter around 500Hz-2kHz and give it a cut to see what muffles the presence of the crowd.
2. Once you have that in place, try a slight boost at approximately 3kHz to get some high end percussive back into the sound.
You could also try some multiband compression or, if this was recorded with stereo mics for ambient, try some mid-side compression and eq in the center of the ambient and let the far stereo kind of come in clearly.