Speech output in the 8 bit era

Jeff Birt takes a look at digitised speech on the C64 in this first of a series:

Skip to the 20min mark for a bit on Forrest S. Mozer, a physicist who figured out - and patented - ideas for low bitrate speech. Links in the video description, including

Here’s the calculator mentioned (it’s for the vision-impaired):

A bit of reverse-engineering of the TI chip inside.

Of course, there’s an Acorn story here too:
Customizing SPEECH! by Superior Software
with in-browser demo.


My favorite all-software C64 speech algorithm:

  1. Record audio on a compact audio cassette
  2. Playback and read in via Datasette, signals will be interpreted as binary value relative to the threshold
  3. Record the stream of ones and zeros
  4. For digital playback, just switch on and off the volume, the SID will produce a short click suitable for low-res, but still understandable audio synthesizes.
    (I think, this could even works in BASIC and simple input# for digitizing. It certainly works for live play-through via BASIC.)

You could build a simple phoneme library from this and advance from there.
It’s charming in its simplicity and for utilizing the on/off click artifact of the SID instead of any of the advanced features of the chip. Since any transition from a previous zero state to a non-zero value will produce the click, you can just push the respective bit to the volume register. Nothing is faster than this.
(Since, as mentioned above, it works for live play-through in BASIC, there should be more than enough time left to retrieve audio data from memory and manage phonemes in assembly.)

1 Like

I recall speech output before the C64. 1979ish on the Apple II - which had a single bit output that used a pure PWM noise algorithm thingy to play back 'samples.

(The original) Castle Wolfenstein was a notable game that used it too.

Memory was key though - those tiny little speech fragments did take up lots of precious RAM!



I think this technique of using the cassette as a simple 1 bit ADC was discovered by a lot of tinkerers including myself circa 1981 and on the Apple II. I remember being a kid and discovering this. Certain sounds like Dizzy Gilespie’s trumpet sounded very good but other things like speech wasn’t so great. Another discovery was that if you placed your finger on the Speaker you could improve the sound quality a great deal. Later I would come to realize that I was effecting a crude low pass filter by using my finger on the speaker cone. It’s intuitive since it tends to muffle the high frequencies when you do that. It was pretty awesome at the time and my father didn’t think it would be possible and yet I was playing digital audio out of the Apple II speaker…

1 Like

The experiment wasn’t without risk, but I discovered in the Summer of '82 (or maybe '83) that the output transistor for the speaker on my ][+ was strong enough to drive a 12" woofer unassisted. It was quite loud … loud enough to impress me and my friends while we played “Bandits”, and also loud enough to annoy my dad while he tried to watch TV downstairs at the other end of the house!