Board Thread:Questions and Answers/@comment-83.134.169.248-20150709083510/@comment-28035646-20160719005551

Yes, I've played around with this idea quite a bit. What I've done is stack at least 3-4 different voicebanks for each part (i.e. for a full SATB you'll have 12-16 separate tracks in the Vocaloid Editor). You'll probably get better results if you use Vocaloids in different languages. For example, I use Yukari, Maika, and Luka English.

In any case, the goal is to try to make each Vocaloid track sound like a unique singer as it would be in a human choir. So, the most obvious way to do this is to set unique parameters for each track (most importantly Gender Factor and Brightness). Humans can't sing perfectly in tune or perfectly in time, so most of the tracks should be pitch-shifted up and down a few cents, and each track should be staggered with a delay plugin, so that there's maybe a 15-45ms difference between each track in a part. Some mild saturation/distortion to one or two of the tracks can also work well in some situations. Then when you're done, drown all of this in lots of reverb.

One last tip: you'll most likely need to do your vocal processing (e.g. EQ, compression, and especially de-essing) on an individual track basis because once the waveforms are combined, it's very difficult to fix any of those issues.

This entire process takes a tremendous amount of computational power, but it's worth it, I think.