Re: Echo II sound cards

[mjmahon@aol.com (Michael J. Mahon)]

Ben Yates wrote:

>Thanks, I read the first faq. They really limit the phonemes. The TI
>has 127 "allophones" [a further break-down of phonemes into smaller
>units]  to choose from, and a few "pause" codes. It also has 64
>pitches and a contouring factor or "slope" (to produce natural
>speech). Plus it has built-in vocabulary words (about 300).
>
>
>And the way words are created on the Echo seems to be a hassle.
>On a TI, a simple PRINT #1:"HELLO, HOW ARE YOU?" in TI Basic will say
>that phrase (if TE-II is plugged in). And using "Input" from a file
>opened to "ALPHON" will return the "allophone" string.
>Or, in XB with Text-to-speech disk:
>CALL LINK("XLAT","HELLO, HOW ARE YOU",A$) ! converts speech to
>"allophones"
>CALL LINK("SPEAK",A$,A,B) ! Says phrase. A is pitch, B is slope
>
>Or, in XB CALL SAY("HELLO, HOW ARE YOU") will say it using the
>built-in words.
>
>I wonder why no one has ever tried to interface a TI speech
>synthesizer to any other type of system. I have other systems and if I
>had time I'd love to try it.
>But still someone would have to write a good text-to-speech system to
>go beyond the built-in words.

The Echo _is_ a TI speech synthesizer interfaced to an Apple II,
so I'm not sure what you're getting at here.

The ability to capture the allophone or LPC streams corresponding to
a given text-to-speech translation is a property of the software, not
the hardware.  The only thing the hardware _ever_ sees is the stream
of LPC coefficients (or, possibly, 8-bit digitized sound samples when
in that mode).

There are a number of text-to-speech programs that run on the Apple II,
most of which use slight variants of the Naval Postgraduate School
algorithm.  Most generate phoneme codes to control phoneme
synthesizers, but a couple generate sound sample streams to send
to a hardware or software DAC.  The one distributed with the Echo
card is the only one I'm familiar with that generates LPC streams
for phonemes.

I don't recall if _any_ of these text-to-speech algorithms provide an
interface to capture the phoneme/allophone/sample/LPC stream that
they generate to control the synthesizer/DAC.  That would not be a
very difficult hack, but my guess is that very few saw it as particularly
useful, since text-to-speech is relatively compact, runs very fast,
and has unlimited vocabulary.

Why, exactly, do you want to capture the stream rather than simply
regenerate it as needed?

-michael

Check out amazing quality sound for 8-bit Apples on my
Home page:  http://members.aol.com/MJMahon/