Re: the new voices for nvda


Luis Carlos González Moráles
 

Will try to contact Guillem. I also lost those files when my external hard drive accidentally damaged itself
.

Josh Kennedy wrote:

He may be using 2020.4, latest stable version. In 2020.4, those voices are not there probably because its an older eSpeak ng. I do recommend for future versions of ESpeak ng, to add more modified klatt voices, and along with modifying those klatt voices, add modified tunes statements such as tune s2, e2, c2, q2, s3, s4, s5, and so on and once you apply them to voice variants, it will allow ESpeak to talk or to speak in a similar manner to other much more familiar speech synthesizers such as decTalk, eloquence, doubleTalk, keynote gold, and so on. Or play with the tunes statements and put in several test variants of several klatt voices both male, female, child, some with lower and higher pitches, more or less bass, more or less trebble for a brighter or darker klatt voice, and then perhaps have a short little 3 or 4 question survey to find out which voice variant people like best or just put them up on the extra voices page. All the variants you have in espeak ng are excellent. But if you would mod the klatt voices some more, and really play around with the intonation file adding more tunes statements and assigning them to different variants mostly based on klatt... ESpeak could be and would be greatly improved. The more familiar it sounds to people, the more they will like it. People have grown up with eloquence, decTalk keynote gold etcetera for 20 plus years. But those are based on 25 or so year old code which will someday break down the road. So why not prepare for the future? Espeak I think is the way forward if devs would just take advantage of its features especially modding klatt voices, giving directions for other to modify klatt voices in various ways, and adding more and different tunes statements in its intonation file. I'll give this to espeak, it supports more languages than eloquence, keynote gold, decTalk, infovox230 from 1993 or so it really puts them all to shame with the amount of languages it supports. OK some may need work. But nevertheless they are supported. Several years ago someone named Gillam, on twitter sent me a voice variant and compiled intonations file and when I applied the variant, espeak began to speak, or talk almost exactly like keynote gold. Unfortunately I lost those files a number of years back. But if he could make it sound nearly exactly like a keynote gold synthesizer, then it could sound like anything you wish by adding some tunes statements and playing with the numbers in those statements. doubleTalk, decTalk, dolphin apollo, whatever you like. 

Join nvda@nvda.groups.io to automatically receive all group messages.