MR> -=> Mike Ross said to Robert Sayre
MR> -=> about "TALKING BOOKS" on 12-09-02  21:56.....

MR> MR> What about text to speech software? I remember the SB16 had something

MR> GM> About 12 years ago at a computer show, I saw a scanner linked to
MR> GM> a computer running text-to-speech software, where you could place
MR> GM> a book or other printed matter under the scanner and the machine
MR> GM> would read each line aloud. The program for converting the
MR> GM> scanned images to text must have been pretty good too.

MR> GM> But IIRC the price was in the tens of thousands of dollars. A lot
MR> GM> more expensive than a cassette player.

MR>I just fired up that 486 with the SB16 sound card in it and ran the
MR>speech software to remember what it did. BTW, I had recalled it wrong,
MR>the name of the program wasn't SoundoLE. It was TextLE and the whole
MR>text to speech group was titled Text Assist in the Win programs menu.
MR>Actually it works quite nicely and sounds much better than the text to
MR>speech thing we are familiar from Stephen Hawking in a wheelchair fame.
MR>This Text Assist actually puts some sing-song inflection and some
MR>expression into the text being read.

MR>In fact there is a demo of the ability of the programmed voices to
MR>literally sing songs. It's a bit of a stiff delivery but it does hit the
MR>notes. I wanted to try my hand at writing up a song but the macros are
MR>rather cumbersome in that the words have to be split up into phonemes
MR>such as WE becomes WIY and things like that. Pitch seems a bugger too as
MR>it looks to be tied to the duration but maybe I understood that wrong.

MR>OTOH speech to text is actually quite easy since it comes with a ready
MR>made dictionary and doesn't seem to need tweaking. There are about half
MR>a dozen preprogrammed: voices, men, women, a kid even, and new ones can
MR>be created too. You can change the speed, the pitch, the pauses, and
MR>various other speech components. It actually all works relatively
MR>seamlessly. In fact I find it excellent for a piece of bundled software
MR>which doesn't cost thousands of dollars but then again as you stated
MR>that was 12 years ago. Only thing is how that program converted picture
MR>images to descriptive speech is way beyond this program's abilities. 

I wonder why, if it is so easy. is it so hard for voice
recognition programs to function reasonably accurately???

MR> Mike
MR> ****

MR>... When I was your age, we carved transistors out of wood.

Heck, that's nothing: We carved them from stone!

Jay
---
 þ OLXWin 1.00b þ Freedom: You don't notice it till AFTER it's gone!