User blog comment:Angel Emfrbl/So Miku English.../@comment-4490524-20120317042912/@comment-53539-20120317104645

If you hear something your mind tries to put together what it is saying. This can led to people not realising how bad something is.

The Japanese Vocaloid can do simple sentances with no problems, but long words and cmplex sentances are a lot harder to work with. I've heard Gumi singing where I only understand 4 words out of the song, but this was a while ago.

Reality is (for example) when she says "I'll give you lots of laugh" She isn't actually saying that. What she is saying is something that sound close enough to that and we work out the rest in our heads.

Also, what a lot of folks don't realise is you can use additional software to manipulate the data to tweak phonemes to sound closer. This a complicated thing to explain, so if you don't mind I'm doing going to do that. However... There are voice syns that can take a single sample and create a whole language from it, the results aren't realistic at all though.