|Apr13-12, 06:40 PM||#1|
Why don't AI systems have "good" voices?
Is it difficult to engineer an AI system that actually sounds like a human (in terms of speaking)?
|Apr13-12, 07:41 PM||#2|
I'm not sure what AI and voice synthesizers have to do with each other, but I can tell you that recognizing and creating human speech, different though they are, are both VERY difficult.
It's one of many areas where human's abilities at pattern recognition is WAY ahead of anything computers are currently able to do.
|Apr13-12, 08:33 PM||#3|
Human speech is "inflected" with changes of pitch, amplitude, and speed (relative length of vowels, etc) depending on the MEANING of what is being said.
"JOHN ran down these stairs" (i.e John ran down, but somebody else did not).
"John RAN down these stairs" (i.e he didn't walk).
"John ran DOWN these stairs" (i.e. he didn't run up them).
"John ran down THESE stairs" (i.e. not some other stairs)
"John ran down these STAIRS" (i.e. not down the street).
Once you can get a computer to understand meaning the same way that a human does, speech generation should be pretty simple IMO.
|Similar Threads for: Why don't AI systems have "good" voices?|
|In binary can we have a value with "deci" "centi" "mili" or more lower valued prefix?||Computers||14|
|"engine and control systems" vs "powertrain"||Mechanical Engineering||2|
|curvilinear coordinate systems and "periodic" coordinates||Differential Geometry||3|
|Extending Newton's Law to real life. No more "ideal systems"!||Introductory Physics Homework||1|
|how to create "good" quantum states from "good" quantum numbers?||Quantum Physics||0|