Wil Wheaton posted a wonderful example of how machine reading of text compares unfavorably to a real performance. He performed a small section of one of his stories and played it next to a standard machine reading. The difference was easy to understand. The human performance trounced the machine.
We know that that text to speech technology does not constitute creating a copy, thus cannot be considered a derived work under the copyright statute. Eventually technology will become good enough to replace the time consuming recording process a good audio book deserves.
For some books a machine reading will be sufficient. Automatic reading will improve over time and more and more books will become very listenable using the technology. But some works of art need to be performed. How will that be handled?
A while ago I was reading about Apple’s text to speech engine and discovered that you can add codes to the text stream to modify the how the machine will read the data aloud. It would be safe to assume Microsoft’s text to speech engine includes the same functionality. It would also be safe to assume the two implementations differ in the codes they use. Someone skilled in the arts could use these codes to start directing the text to speech engine, thus creating a true performance. This would never happen in a world with multiple standards.
I propose to assign a Unicode code page to create a similar set of codes that all text to speech system could understand. When a text document is sold to a customer, the codes could be included in the text stream, but made invisible by the Unicode standard. Or they could be sold separately. This will allow reading and listening to the same file.
Authors to not have the right to be paid extra for reading aloud of their text. If automatic text to speech is good enough for you then the authors will not have an opportunity to sell you an additional copy of their work. But using this scheme, authors could go the extra distance to add value to their text and then deserve the extra money.