🚀 go-pugleaf

In <YekpR3200UhWE4tIth@andrew.cmu.edu> Matthew Keith Albright <ma2b+@andrew.cmu.edu> writes:

>Okay, I would like a project that I am workng on to be able to speak. 
>Like maybe 10 - 20 different 3-4 second messages (In perhaps three
>languages).  What is the easiest (cheapest) way to do this?  My ideas so
>far: 

>1) Get one of those phonic-based chips that can say anything, all you
>have to do is arrange the individual phonics.  Good: flexibility  Bad:
>sound quality.

>2) Pre-record the messages and somehow get them into EPROM, using a
>vocoder or something like that.  Better sound quality but less
>flexibility and a lot of memory is needed to store the messages.

>The method chosen will have to be able to be interfaced to a
>microcontroller.  So are either of these ideas valid, or is there a
>better way to do this?  Any ideas are welcome.  FYI, I am leaning toward
>idea #2.

>Thanks in advance, and please reply by email.

>matt
>ma2b+@andrew.cmu.edu

There is a book called "Chip Talk:  Projects In Speech Synthesis", by
D.  Prochnow; Pub. TAB ISBN: 0-8306-1912-7
I guess it has many suggestions you might like to incorporate in your
project.  Also, filter your talk to <3KHz, then use PCM for your second
method. This way you use 16 levels or 4 bits per sample and 2*4*3K=3KByte
per second.  I do not think you need more then 10 minute worth of massage,
so you'll need only 10*60*3=1800 or less than 2Mbyte memory.  You can get
it as two 9*1M SIM cost only $60.00.  10 minutes is a generous massage time,
too.

                       With highest regards,
                            Babak   Sehari.