[sdiy] Texas Instruments LPC Speech Chips

[Ben Stuyts]

Unfortunately NeXT never got a good DSP development environment or SDK of the ground. Pity, it was one of the reasons I bought the thing in the first place. So rsynth didn’t use the DSP, only the 68040. As far as I can remember, it was mostly about writing a small audio driver and solving some K&R C issues.

I can’t remember the software you mention. I put your description through the clanker and it comes up with this alternative:

“
SPASM = Singing Physical Articulatory Synthesis Model, written by Perry R. Cook at Stanford’s CCRMA. It was specifically a NeXT application, and the original distribution was on CCRMA’s NeXT FTP archive as SynthesisTools/SPASM.tar. The old CMU AI repository still has a catalogue entry for it.

The contemporary description calls it: “SPASM, a Real-Time Vocal Tract Physical Model Controller and Singer” and describes an interactive NeXT GUI controlling a physical vocal-tract model. It used the NeXT’s audio/DSP facilities and allowed parameters to be manipulated graphically.  
“

Ben

> On 2 Sep 2026, at 01:44, brianw <brianw at audiobanshee.com> wrote:
> 
> Keep in mind that the NeXT hardware models all have a Motorola 56000 DSP, as well as the m68k.
> 
> I remember rsynth, but I do not recall whether it took advantage of the DSP.
> 
> I just re-read your email, and see that you ported rsynth to NeXTSTEP, so I guess you would remember if you used any of the DSP resources.
> 
> Brian Willoughby
> 
> p.s. I'm still trying to remember the name of the NeXTstep software that modeled the human vocal tract, with a graphical depiction that you could manipulate on the screen with the mouse in real time to synthesize vocalizations (but, alas, words were a bit too complex to recreate with mouse drags). I think it was called GRASP. Something like Graphical Representation Audio Synthesis something-er-other.
> 
> 
> On Sep 1, 2026, at 5:03 AM, Ben Stuyts wrote:
>> That reminds me of an old project I ported to NeXTSTEP back in 1994. It was called rsynth, an early open-source text-to-speech system that converts text to phonemes. It generates speech using a Klatt-style parametric formant synthesizer.
>> 
>> So, certainly not a discrete solution, but it was very small and fairly intelligible. Reminded me a bit of the Votrax and TI voices. I wonder if it could run on a modern microcontroller at reasonable speed. My NeXTcube had a 25 MHz 68040, and it wasn’t a real time experience...
>> 
>> I see it has been updated as recent as 2005 ;-) https://sourceforge.net/projects/rsynth/
>> 
>> Ben
> 
>