Crystal clear speech -and in any language
THE Voxbox is yet another device to allow your computer to talk.
But, unlike the Acorn speech synthesiser, its vocabulary is unlimited.
Other speech synthesisers claiming this use a system known as
allophones and, although they also offer an unlimited vocabulary,
it is not very clear.
The Voxbox, however, produces crystal clear speech in any accent
or language you require.
It uses a technique known as Delta pulse code modulation to
effectively make a digital "tape" recording of any sound
you care to input into its microphone.
To create a word or phrase, you speak it into a microphone.
It is then digitised and stored in memory and can be replayed
so you can hear if it is correct.
Several words or phrases may be defined in this way until you
have used all the memory reserved for speech data. You may then
replay the phrases or words, strung together in any sequence you
like.
Along with each section of speech is stored a string variable.
This is usually what is said, but can be anything.
The memory containing the speech data can then be saved as a
file on disc and incorporated in your own programs.
The problem with this type of speech synthesis is that it takes
up large amounts of memory, although this is eased in The Voxbox
by using delta modulation.
This is where the size of the digital sample is not stored,
but rather the difference in size between the last sample and
the current one.
Also, the rate at which samples are taken determine the quality
of the final speech. The faster the rate, the better the speech.
But this uses up the memory faster and you cannot get as many
words stored.
The software provided with The Voxbox allows you to choose between
one of 10 sample rates to enable you to get the best length and
quality trade-off for your application.
These give you approximately six to 16 seconds of speech, using
the maximum amount of memory available.
The quality at the fast rate is very good indeed, almost as
good as a cassette tape recorder or a telephone.
As the rate slows down, the sound gets increasingly muddy. However,
it never degenerates into the awful "computer speech"
much beloved of .film and television producers.
Once you have built your voice files, they may be incorporated
into your own Basic programs.
There is a small machine-code module which allows you to do
this and on the disc I received was an example of how it can work.
The program is called "GREET" and gives you a choice
of five different languages to try. After selecting one, the screen
displays the word of greeting in that language and the Voxbox
says it.
You are then requested to imitate the word and your attempt
is recorded. As a final act of humiliation, the original word
and your attempt are played back to you.
My particular favourite language was Australian. Apparently,
the Australian for hello is G'DAY.
The Voxbox really does require a set-up with discs to handle
the large voice files you create. I can see this being especially
useful for teachers who want to write programs which ask the children
to spell particular words.
The bank of words can then be individually tailored for the
age range or specific difficulty of children.
In fact, you do not have to stick to speech. Any sound can be
recorded and played back in your application.
All in all, an excellent product, reasonably priced, and one
that the editor had extreme difficulty getting off me to return
to the manufacturers.
Voxbox is produced by Multiplex and costs £89.
Mike Cook
