Voice on Spectrum (2Max Romanoff)

Andrew is that you?
...why did you tell All about Voice on the Spectrum (2Max Romanoff)?

AS> Yes, but how much memory will it eat up... how many sounds are there
AS> in Russian? And 2 kilos for each sound?
AS> And then, from them it will be possible to obtain by transformation
AS> only those voices that you get on a tape recorder when
AS> accelerating/slowing down the playback.

a... you're talking about the SYNTHESIS of voices... well, here you need to work with phonemes...
still without sampling is indispensable.

AS> Well, will it be nice to listen to such garbage in toys?

better than a synthesized voice...

AS> half-duplex and replaced German with Russian-English ;)

there were devices for Sinclair. speaking. a couple of games supported them

AS> But if only someone could explain how to synthesize speech with different
AS> timbres. And so as not to consume all the resources of the z80. If this
AS> is impossible, then at least what are the most commonly used
AS> operations required (and what should be the time for their execution),
AS> that is, not just say “impossible”, but explain why it is impossible.

I explain why in any case it is impossible to gobble up all the resources: due to
the absence of audio DMA, the percentage must output
with a certain period sample on a DAC or beeper. this time. two - this still needs to be counted.

a simple option is to count in memory, then lose.
for perverts -count in real time.

in any case, the output will eat up 100% of the time.

AS> So you threw away “samples” briefly, so give an example of some thread
AS> synthesis, (yes, by the way, making a library of sounds yourself “by hand”
AS> you’ll be tormented, you need a program that voice sounds are pulled out,
AS> but this is automatically even more complicated than synthesis, just like
AS> the transformation pcx -> txt is more complex than txt -> pcx) )

the simplest thing is to sample syllables.

AS> If you did this, explain what it all means spills out in memory and
AS> processor load.

in terms of load - a lot, in memory - depends on the specific implementation.

AS> Doesn’t anyone want to study languages? Yes, and the music in the demos with vocals was
AS> more fun. As for educational programs and toys, I’m silent.

As long as the Spectrum is 1 percent, there won’t be any vocals.

AS> In a word, help the teapot, whoever can! I think these questions are interesting
AS> not only to me. It’s not in the FAQ (this surprises me every time)

I don’t know, but such nonsense (sorry, but it’s true) seems to have never occurred to anyone..

AS> In the latest versions, the voices were recorded from a microphone - that’s for sure!

where else could he have recorded them? Read lips, or what?! :)

AS> then they made a digital answering machine and voice memory - it’s clearly based on
AS> the sametechnologies: digitization and packaging, because without packaging you
AS> you won’t be able to record much in 64K.

you will. you can record from 40 seconds of sound.

AS> On ISKRE-226 (April 1, 1983), when we stupidly digitized a 10 kHz ADC,
AS> only the phrase “And your whole back is white,” said by a normal
AS> tempo.

in those scarves, most likely there is only 1-bit sound, when you had 8-bit.

Good Luck...

... God made wrinkles to show where smiles have been.