Posted By
Gaia
on 2017-02-05
15:20:09
|  Re: Creating new Speech Data for the V364
@George: go for the 'SHORT' type. The T6721A operates with 15 bits internally. You also need to match the bit resolution (9 bits, but go for 8 bits), the output frequency (8 kHz) and the bit rate (96 bits per frame like for most words as discussed, since this is the only known bit format). Actually the bit rate does not really matter, just make 12 new PARCOR parameters - 1 pitch, 1 energy and 10 filter coeffs - per each frame (every 20 ms). You can scale them back to the appropriate bit format later. You also have to make sure you set the frame interval to 20 ms (ie. 50 Hz). In between frames the parameters will be interpolated by the T6721A.
@MMS:
It's not the tooling which is/was the bottleneck but the capabilities of the PARCOR speech synthesis. It is modelling the human vocal tract in an extremely condensed format with low output frequency. It is 9 bit which is wonderful, so I believe where we could use it is rather playing music with it, rather than vastly improving the speech quality. Remember, the Magic Voice demo could even sing.
@JamesC:
You can already do that 
SAY"I":SAY"AM":SAY"THE":SAY"COMMODORE":SAY"V":SAY 3 :SAY 6 :SAY 20:SAY 4
EDIT: Yape should play uncompressed as well. So I need to have a look, most likely it's a bug. I haven't touched the code for more than 10 years now. Stay tuned.
EDIT2: OK I checked and ABC uses the 48-bit which VICE does not support either, so I am a bit confused now how it could work for you.
|
