Advanced TI Speech

[cph1776]

Hello all,

 

In early 2016 I was discussing various ways to use the TI Speech Synthesizer. My goal was to (eventually) be able to convert a .WAV file into a format that could be read by the TMS5220.

(Like QBOX, but with less hassle). Lots of discussion on SPEECODER, STANAGs, and SOX. The thread is at https://atariage.com/forums/topic/251554-speecoder-on-win994a/

 

Not too long after that, I picked up another part-time job, and ended up writing three (non-computer related) journal/magazine articles, taking up much of my time and brainspace.

TI speech, unfortunately, got shunted off to a siding.

 

Now that things have quieted down a bit, I'm revisiting my investigation into the speech synthesizer. One project, if anyone hasn't beat me to it yet, is to extract speech data from TI's text to 

speech programs ("SPEAK", "XLAT", etc), and save it to a file. (Note that the input expected by "CALL SAY" is different than the output of XLAT--CALL SAY wants TI-LPC data, XLAT produces a

coded phoneme string.)

 

There will be other studies in LPC formats, bit manipulation, phonemes, formants, and I may even tackle the math that generates the reflection coefficients. But for now, enjoy this Extended Basic program

that makes the TI sing! (feel free to scroll ahead to about 1:30 to hear the actual singing...) Let me know what you think. 

 

 

 

[Nick99]

An x-mas song, nice!  

Also fun to combine music and speech in that way, I can´t recall ever heard TI singing with music. 

Looking forward to follow your progress on the WAV to speech program. ?

[ti99iuc]

a very nice good job!

 

[HOME AUTOMATION]

I must admit to being somewhat fixated by early speech synthesis techniques and LPC. 

 

Classic99's debugger can capture data from >9400(very useful).

 

I as well, confess to having, breached the surface a little...

 

#25

 

...and are you aware of this:

 

examples

 

I would like to expand these vocabularies or phonemize them.

[cph1776]

4 hours ago, HOME AUTOMATION said:

I must admit to being somewhat fixated by early speech synthesis techniques and LPC. 

 

Classic99's debugger can capture data from >9400(very useful).

 

I as well, confess to having, breached the surface a little...

 

#25

 

...and are you aware of this:

 

examples

 

I would like to expand these vocabularies or phonemize them.

Hi. I downloaded the tts2lpc disk and played around with it. One issue is I'm not quite sure when to turn the buffer on/off before extracting the lpc codes. I thought I got something from CALL LINK("LPC") but when I ran it through

CALL SAY, it sounded garbled...

 

The talkie source code looks *very* interesting. Seems easier to follow than the source code in SOX, for one thing. I'll see if it gets me closer to the "holy grail" of converting WAV files to TMS5220 LPC...

 

[HOME AUTOMATION]

The idea of using OFF is just to keep the buffer from being overwritten(if desired).

ON is to clear the buffer.

Did you run PHRASE and use FCTN 4(BREAK)? ...see PAGE 2 of the thread.
Are you on real hardware or EMU?

 

I imagine TALKIE doesn't crash. I haven't tried it though. Curiously simple!

[cph1776]

10 hours ago, HOME AUTOMATION said:

The idea of using OFF is just to keep the buffer from being overwritten(if desired).

ON is to clear the buffer.

Did you run PHRASE and use FCTN 4(BREAK)? ...see PAGE 2 of the thread.
Are you on real hardware or EMU?

 

I imagine TALKIE doesn't crash. I haven't tried it though. Curiously simple!

 

I just tried it now. It works fine. (I had difficulties getting some of the programs to load last night for some reason....) 

 

This is a jump ahead for me (I was preparing to disassemble the TI programs to find out where the LPC codes were stored!)

 

I might try adding a feature where it saves the LPC strings to a disk file, or even BASIC program lines in MERGE format....

 

I am currently using MESS .0159.  (I have MAME .216, installed, but I haven't tried to run TI99/4a emulation there yet)

 

Talkie looks like source code for the Arduino, which I haven't played with yet. The algorithms in the source code might be handy for another project, though.

[+mizapf]

1 hour ago, cph1776 said:

I am currently using MESS .0159.  (I have MAME .216, installed, but I haven't tried to run TI99/4a emulation there yet)

Remember to get the suitable ROM dumps from WHTech (folder System ROMs/MAME). Here is a list what happened in the meantime for the TI family: https://www.mizapf.de/ti99/mame/changes

 

[HOME AUTOMATION]

I was more desperate to capture the data stream than I was to determine the means of operation. Around here... it does make it feel like I cheated or something!

 

Sometimes just a few short words generate enough data to overflow the string length limit of 255. I worked on an alternate method to move the data within XB, but lost interest at the last minute.

 

Glad that you got it to work.

 

I mainly made it because of performance issues, I wanted proper vocabulary, better speed and I needed access from running machine code. I wonder if TE IIs virtual "SPEECH" device would even work outside of the BASIC environment.

 

I believe TALKIE was developed on TEENSY 2...
 

https://www.pjrc.com/teensy/

[tr1k1]

On 12/21/2019 at 10:37 PM, cph1776 said:

Hello all,

 

In early 2016 I was discussing various ways to use the TI Speech Synthesizer. My goal was to (eventually) be able to convert a .WAV file into a format that could be read by the TMS5220.

...

[tr1k1]

Sorry, it wouldn't let me put any text in the post above!
There is a fairly new project to convert .WAV (and other) to TMS5220 (and others).
https://github.com/tornupnegatives/TMS-Express

[pixelpedant]

Neat.  Looks like, indeed, it comes with TMS5220 reflection coefficient values (K values) in source, and these would consequently need to be modified for the sake of optimal use with our TMS5200.  Just as previously, with BlueWizard.  But it's great to have another tool.