Vocaltract on Hydra? Probably asked before.

VIRAND

Can anyone point me to a clue how to run and hear the "seven" or "monks" demo on Hydra?
(and then tinker with the vocaltract)

I tried probing pins with an audio amp looking for the sound. Maybe I missed it.

What else is different in the program besides the sound output,
and can it be edited to work? I just started to tinker with it so it may be
a while before I figure it out alone.

Also, I have "software speech synthesizer" development experience from the 1980's and 1990's.

JT Cook

There are two things you need to do. Change the clock speed and multiplier, and remap the audio pins from the pinout from the demo board to the hydra pinout.

Mike Green

Also, the Demo Board has stereo sound while the Hydra has monaural. I believe the vocal tract software only produces one sound stream, but there is a localizer driver that generates stereo sound from a monaural sound stream and position information ... that won't work on a Hydra.

Baggers

Mike,·I've already converted the singing demo to·the Hybrid which is monaural like the Hydra.

I've attached the files source and binaries for Hybrid and Hydra.

VIRAND enjoy [noparse]:)[/noparse]

VIRAND

Thanks... sounds beautiful although it was noisy at first, even on another speaker.
I was able to make it less noisy by finding a bug that doesn't turn off the voices between measures
(somehow resulting in background white noise) ... a vtract[noparse][[/noparse]state]:=0 instead of ~ seemed to help a lot.
And also by changing the echo constant to 000_100_000_000 it got even less noisy.
And when I tried to change the formants to higher ones like "oo" and "ee",
it got distorted so I had to comment out one of the four voices. Still a tiny whisper of noise though.

The noise is not heard in the demo mp3's.
Maybe the stereospatializer and clock speed are a problem on the hydra;
maybe I just need an EQ on my speakers.
Maybe I'll compare the proto and hybrid code to the hydra code for clues.
Sorry, just talking to myself.

The vocaltract has awesome potential. I want to sing a song with it.
If I write anything good on it I'll post it. Thanks again.

Baggers

Yeah, that would be cool to hear a song on it [noparse]:)[/noparse]
And it does have lots of potential.

Try pointing one of the output's to the debug LED, it might be because both channels are going to the same pin.

mpark

VIRAND said...

I was able to make it less noisy by finding a bug that doesn't turn off the voices between measures
(somehow resulting in background white noise) ... a vtract[noparse][[/noparse]state]:=0 instead of ~ seemed to help a lot.

I'm getting the noise too. I can't find "vtract[noparse][[/noparse]state]" anywhere though. Which file is it in?

VIRAND

It's in the source code file, in the main loop that sings, near the end of the loop.
By changing it, the program sings forever instead of ending with endless noise.

VIRAND

FYI+FYE: Noise is ALL GONE now! ...

I downloaded the SEVEN demo ... of course it was silent.
I Changed the stereospatializer parameter from 11 and 10 to 7 and 0. Noise gone in either demo.
This apparently means that the sound will come out of just "Pin 7" instead of 10 and 11.

Before that I made it two 7's, that makes noise. (7,-1,0,-1 = nice 7,-1,7,-1 = yuck!)
...babbling on about it...
The "knobs" parameter above that represents an echo delay line and can be left alone.
Setting "knobs" to all 000_000_000_000 sounds "ok",
but all ones sounds like many rodents trapped in a grand piano.

I also changed the CONstants to xtal1+pll8x and 10_000_000 for the crystal timing for Hydra.

SevenDemo is interesting for trying to make it sing something else...
so far like "never seven heaven", based on obvious comments in the source.

It sounds like untrained People singing if you lower the vibrato rate moderately.
Lower it too much and it sounds Robotic but still a little better than Speak n' Spell.
(I only tried that in the other demo given a few comments above, haven't yet found it in the Seven source)

Baggers

Good work so far, I guessed it'd be both channels pointing to the same port.

Thanks for the update. [noparse]:)[/noparse]