[GTALUG] Actual ttyS0 MIA

[Stewart C. Russell]

On 2020-05-11 1:57 a.m., Karen Lewellen wrote:
>
> Actually, that it has a default makes a grand deal of sense.  After all, 
> how often would a person have say 4 of them running at a time?

Um, me, aka the outlier? Right now I have a CP/M machine on ttyS0 (so 
not USB), the pen plotter on ttyUSB0, the PDP-8 on ttyUSB1, a 
microcontroller doing Python things on ttyUSB2, Arduinos occasionally 
connecting as ttyUSB3 or ttyACM0 depending on model ...

... and I don't even *like* serial devices! Yet I seem to be the Spiders 
Georg of serial ports despite that.

> ...
> Wait!  you built a $600 dectalk USB unit in your garage?

In my front room, for about $100, but yes.

> Does it provide all 9 voices?

Yup.

> Can you make it sing?

Kinda. Most DECTalk songs - and there are *lots* - are written for v4 of 
the hardware. The board I'm using gives v5, so some of its timing is off.

> ... I prefer to have a spare, for  slightly less than a thousand, and 
> often find them sold for around $1,500 even today.

They're still expensive because "DECTalk Voice" is a big thing in 
electronic music. Assistive tech is always expensive, but couple with 
ebay competition from music nerds and Hawking fanboys, the price goes 
through the roof.

> ...
> Stewart, your comments about speech quality has me wanting to ask 
> something about speech in  Linux, which for the record is reprehensible, 
> even if free.  IBM created a voice known as eloquence ...

I'm not a TTS user, just someone who geeks out on the technology. So 
much of TTS seems to be very proprietary, thanks to a few IP harvesting 
companies. Eloquence seems to be an engine that can run under a screen 
reader framework, and Orca maybe the only game in town for Linux as a 
screen reader. Licensing actual voices for IBMTTS/Eloquence seems to 
cost money: not much (maybe $5-10 per voice) but with strict 
noncommercial terms.

I was hopeful that the SVOX Pico TTS system as used on Android would 
gain more traction, but it's shipped with limited voices and tools. 
While the code is open source, the parent company was (surprise!) 
snapped up by Nuance, so we'll never see it go beyond its 2010 capabilities.

cheers
  Stewart