Two AI images from the "flamin' arena" set generated with the qwen image model.

@Tamasg Quick question for you. Is there a way a programmer can test an NVDA plugin without having the NVDA screen reader installed, or is that pretty much a requirement?

A demo I made of the DECTalk 3.0 a while ago, and sent to Aksel who runs the dectalk.nu site. I have a working board (it's just the bare PCB board) and I'm thinking about getting rid of it. Any takers?

@arroz LLMs are NOT random content generators. That is false. The LLM output is based the user prompt. Seems to me you don't know how to prompt correctly.

Apparently, there was an obscure Windows port of the Macintalk speech system called WinTalker. I was able to successfully build it and run it.

@Tamasg Yeah it's good stuff. It's a good place to start.

@Tamasg You might want to check this file too. It's got some good stuff on target adjustment.
https://github.com/dectalk/DECtalkMini/blob/dectalk-develop/include/p_us_st0.c

@Tamasg Have you considered adding formant target readjustment rules to your program? That's something DECTalk did, especially for back vowels after alveolar sounds.

@Tamasg Why so quick to give up?

@Tamasg that sounds right. closures happen quick.

@Tamasg Same timing.

@Tamasg DECTalk actually has locus times usually about 35 ms to 50 ms. Longer than what you're suggesting. From my testing, between 30 ms and 70 ms is most natural. 50 ms is most common.

@Tamasg A suggestion is to add a locus timing parameter in the formant table. Some vowels have longer or shorter locus times.