[DECtalk] Question About Formant vs Natural Speech

[Don]

On 6/19/2022 7:36 AM, Karen Lewellen wrote:
> Personally, I would add that the screen reading program can also influence how
> a synthesizer voice sounds.
> I have used the dectalk synthesizer of my reading edge for years, finding it
> sounds different, though be it slightly,  when the program in question is say
> vocal eyes, as   opposed to say business vision, or tiny talk.
>   might say the same between use of say a dectalk express or internal card, and
> the edge as well.

That could be a result of how the synthesizer and screen reader interact.
E.g., there are no limits (in the English language) on the length of a
particular sentence (assuming it is grammatically correct; if you throw
out that requirement, then there TRULY are no limits!)

But, the synthesizer needs to set aside resources to process each utterance.
Should it accommodate 40 characters?  80?  180?  How much "context" does
the synthesizer need to differentiate between different ways that *it*
might pronounce a given string of characters?

I previously used the example, "Ben went home*".  What if it had been
"Ben, the owner of that fabulous Bentley parked under the Chestnut tree
in the front yard on the south side of the house, went home?"  Just
how large of a buffer should the synthesizer set aside to process this?

> Only speaking personally of course,   but I dare say a bit of human
> interpretation  might be factored in too.  comparative to those who can hear
> the difference between an mp3, and actual  quality musical playback.