It seems to me that it should be possible to “push” the sound of a crowd talking through a voice synthesizer. Shaping the noise (which strikes me as essentially schwa-ish) into any vowel of the english language, and sticking it together with consonants to form all the phonemes you’d ever need.
Questions are: 1. Is this possible?
2. Would this be recognizable, and understandable, as speech?
3. Anybody try this?
Sometimes at stadiums, a whole bunch of people will be cheering using the same words. It sounds like a crowd noise, but with a voice.
Ah, they’d just all turn out to be saying “nadder nadder gramish gramish” or “mashed potato rubarb rubarb.”
[inside acting joke]
I thought it was “watermelon watermelon” or “peas and carrots.”
Well, I come from the “nadder nadder gramish gramish” school of acting.
I think it would be possible.
The question to me is: would the sound of a crowd work better than white noise as an input to the synthesizer?
I could try it. I don’t have a speech synthesizer but I do have two vocoders. Vocoders analyze the spectral content of a source sound and apply that pattern to another sound - like a synthesizer, white noise, or a recording of a crowd.
It might take me a couple days to get results. It depends on how long it takes me to get a recording of crowd noise. And I don’t have my own website, so if I want to post soundfiles I’ll have to ask someone to donate a little space.
There used to be a mynah bird living in a pet shop (Lucky Dog) in Berkeley who would articulate the sound of the crowded store. you could even pick out some individual words. Is that what you mean?
No? 
Peace,
mangeorge