I would make several copies (depending on how large an audience I was looking for). slightly change the size of each of the copies by control dragging the edges, slightly offset the copies on the timeline, maybe change the eq on each clip, add a little reverb to some, change the level on others, in short make each clip sound slightly different and occurring at a slightly different time.
What I would do is duplicate each one a few times on separate tracks and change the pitch up or down a level and move them off by a frame or two. Then I might add a little bit of a crisp/bright reverb depending on what I was going for.
why not layer the recording ? ie, record multiple takes of them saying the word, and ask them to vary their voices each time.
One more thing -- if you've got a good recording space, you probably want to vary the distance between the voice and the mic to place everyone in the soundscape (as a real audience would be)
Your suggestion is very good.
I've tried doing something similar and trying to get dry recordings to sound like they belong in a space, especially a large space is no trivial task. In a large space filled with people distant sounds become muffled as all the bodies in the space absord sound.
Tried that, fine for a single sound source at a given point in an empty space. Multiple sound source (people talking and shouting) in a large venue is another matter. Even trying to add a few extra voices into that space has so far eluded me. Maybe I'd have a chance if the production had recorded impulses from multiple places on the day but they'd have needed the crowd to sit still and keep quite. Sadly none of that happened.
Based on my rudimentary knowledge of how multiple sound sources interact in a complex space I have no clue as to how it can be done to build a large crowd saying the same word in a large venue.
You say "audience" rather than "crowd" so I presume the people are in an auditorium, and therefore you want to include the room effect. Nearby voices will be loud and have little reverberation. Distant voices will be fainter and include reverb plus some frequency dependent loss, probably the low end. If all speakers are actually saying the word in unison, then the speech from distant speakers will be delayed.
So far as the speakers are concerned, you will want to introduce some randomness in the timing, pitch, timbre, etc.