1980's speech synthesis chips | Hacker News Reader

1980's speech synthesis chips

eevblog.com

56 pointsspeps20 comments

Chips? Luxury! In the mid 1980s the ZX Spectrum only had a 3.5MHz Z80, a sound chip that could only beep, and synthesis had to be done in hand-optimised assembler! Uphill, both ways!

https://youtu.be/3qqfNgA8_OY?t=1m58s

I think this one from Apple II is just recordings not synthesis

https://m.youtube.com/watch?v=9sptDXptWeM

Here's one that might be synthesis

https://m.youtube.com/watch?v=Y3QV_2hq8qU

Apple II was a 1mhz 6502 with a click speaker

Oh man those were the days. If only I had the moxie to stuck with it, I might have made something of my life.

It was not so much sound chip as a GPIO connected to a speaker.

C64 had a 1mhz 6502 and had way better speech in the form of SAM (software automated mouth). The c64 had a real sound chip (SID, arguably the best of its peers), but SAM did not use it except as a PWM output - it did the formant synthesis in software, and used the SID only to output the resulting 4-bit waveform.

(Which is 3 bits more than the ZX spectrum had... Or the PC effectively until 1990 or so when sound blaster became ubiquitous)

Does anyone have any links/info on the algorithm used in SAM? I remember playing with it as a kid on my friends C64 and then trying to figure out how I could make my own version of it on my obscure (in my part of the world, at least) SpectraVideo https://en.wikipedia.org/wiki/SV-318 using only BASIC "Play" commands. Needless to say, I failed miserably and have been wondering about SAM ever since!

https://github.com/s-macke/SAM has everything you need to make a SAM clone. Does your SV-318 still work? :)

I still have my BBC Computer from then. It has a Speech Rom [1]. It is the voice of the first ever BBC tv news reader Kenneth Kendall. He was also in 2001: A Space Odyssey.

[1] says the chips are a TMS5220 and a TMS6100. It was pretty good fun making the computer speak using BBC Basic.

[1] http://www.retro-kit.co.uk/page.cfm/content/Acorn-Speech-Syn... [2] https://en.wikipedia.org/wiki/Kenneth_Kendall

As a legally blind person who started elementary school in the mid 1980s, the speech synthesizer that I associate most strongly with that time period is the Echo, installed in an Apple IIe. Here's a video demonstrating the Echo with an Apple II+:

https://www.youtube.com/watch?v=sbZuBhT7HK4

The MESS emulator can emulate the Echo, as well as the Apple IIe itself. I spent some time playing with that configuration on my PC, for a bit of nostalgia.

I was shocked to learn years later that in the 1980s, the state of the art in speech synthesis was actually DECtalk:

http://www.festvox.org/history/klatt/part35.au

But DECtalk was probably too expensive for a public school system at that time.

By the way, that clip comes from a record produced in 1986 by Dennis Klatt, the late creator of DECtalk. It's a fascinating chronicle of the history of speech synthesis:

http://www.festvox.org/history/klatt.html

If you want to play with these things in software, Plogue make a very detailed emulation of many of these chips in their chipspeech synthenesizer. It's nice because it has all the chaotic weirdness of those old algorithms but you have complete musical control, enabling you to make them sing.

https://www.plogue.com/products/chipspeech/

Thanks for posting. I always wanted to emulate the original 1930's vintage Voder speech synthesizer in software. I wonder if anyone's done it yet. The Voder was not computer controlled, it had to be played by an operator, much like playing a piano. http://120years.net/the-voder-vocoderhomer-dudleyusa1940/

That's what got me interested in speech synthesis as well. It seems very possible to do in software, the patent is quite interesting to read : http://www.google.co.uk/patents/US2121142. Not sure when it expires these days...

As an alternative to locating old chips, it is possible to re-create old style speech synthesis in software with a little effort. Google Linear Prediction of Speech and you will find many leads. Linear prediction speech synthesis was a US vocoder federal standard at one time, standard FS-1015. It made everybody sound like Speak'n'Spell.

I had an SPO256-based board for my Amstrad CPC! It was programmed using allophones (datasheet[1]). AIUI this is a deprecated way of doing speech synthesis, and indeed the results were quite rubbish. How do modern speech synthesizers work?

[1] http://www.futurebots.com/spo256.pdf

There's a pretty cool large-size modular unit with the Votrax SC-01 chip in there. The video at the bottom has a demo - http://www.noisebug.net/site/largemodular_syntheticsoundlabs....

Long ago I learned the peculiar way the early speech synthesizers pronounced the phonemes and decided that it was just like learning any other accent, so I studied and practiced.

I can still do it today. The younger folks haven't heard it as much so they don't understand the reference, but the people my age freak out when they hear it.

The one you could buy for the TI99/4A was fun. It had BASIC commands and you could do a lot of adjustments. I programmed one to say Dakota words better than I could. Still a bit ticked I could get a computer to do gutturals better than I could.

Does anyone else remember making Speak and Spell say rude words? :-)

Back in 1984, I was at Disney world with a school trip. They had an exhibit that looked like an arcade game that had a keyboard. You typed on the keyboard and it would say what you typed. We almost got kicked out of Disney world for making it say rude words. I remember exactly what the employee said. "If I hear one more cuss word out of that thing, you are out of here."

We had SAM for our Commodore128/64 and once you loaded the program, you could write BASIC programs that used it to include speech. Pretty sure I spent way too much time figuring out just the right combination of phonemes and numbers (used to denote accented syllables among other things) to make it swear convincingly.

Reply on news.ycombinator.com