Quote:

AI is lagging far behind in my opinions as regards voice recognization. In this day and age you would expect software to take a sample of a person's voice (say an hour long) and then be able to speak in a natural voice of that person any type of text that can be input to it.
musiclover



My reply will be somewhat OT, but will address the above:

I suspect that recording a voice, and using the recordings to make the voice say something natural sounding will take some time yet.

Perhaps the simplest instrument to sample/reproduce is the piano, and although there are some very good recordings, (samples of each and every note at various velocities), when the "bits" are put together, one can almost always tell the sample rendering from the real thing. There is no continuity or connectivity between the notes.

The problem is far greater with the human voice (or wind instrument or guitar) because there are variations in tone, pitch, and timing that make the real thing extremely difficult to emulate. Certainly much more difficult than emulating the piano with samples.

It may be possible to emulate from a theoretical point, but in practice, it's very daunting to say the least.

It might take a whole lot more than artificial intelligence.

Glenn