Re: I take it there have been no improvements to the flawed espeak ng in 2017.4?


Sharni-Lee Ward
 

Okay, I installed NVDA 2018.1, after creating a portable copy in case it became too much to bear.


I don't like Eloquence because it isn't as crisp as Espeak. I have compared it to A.M radio in the past, and Espeak to FM/CD Quality sound. Also, it has only eight variants as opposed to Espeak's forty-something (I just realised my lovingly customised alternate female variants have somehow been broken by more recent changes which makes me sad, but I usually stick with Linda anyway).


It's still doing the things that annoyed me in December, but I'm willing to try and see if I get used to them.


I don't expect a program to be perfect. I don't expect a synth to be perfect. I would like to hope that people don't break something that was perfectly all right in a previous version, however.


And as for inflection, I have it set to 100% (GAH, it does the thing with percent....), which makes Eloquence sound overly chipper, another reason I don't enjoy that synth as much. By flat, I mean the voice does not use perceptible inflection in the cases of question marks, commas and exclamation points. This should not be something a blind user has to put up with when it comes to synthesised voice. How is one supposed to know if the statement is yelled or spoken calmly if the voice doesn't give it away? How emotionally invested can they be in a story if there is no noticeable inflection for something as basic as exclamations and question marks?

 


On 25/03/2018 4:54 PM, Gene wrote:
I need to correct one statement.  You can't, as it appears, change inflection in Eloquence with NVDA.  I don't know what you consider flat.  But if you give Eloquence a chance, you may find that the amount of inflection it uses is not bothersome to you.  You may want an exaggerated amount of inflection.  I don't know how you have E-Speak set.  But you may find that, if you give Eloquence a chance and get used to it, you may find it generally superior over time. 
 
Gene
----- Original Message -----
From: Gene
Sent: Sunday, March 25, 2018 12:50 AM
Subject: Re: [nvda] I take it there have been no improvements to the flawed espeak ng in 2017.4?

Also, what's wrong with Eloquence?  Have you tried it  And as I recall, you can adjust inflection with it.  Don't use the Brittish English voice.  It's obviously programmed by Americans amateurishly trying to imitate a Brittish accent.  Use the default American English.  It pronounces a great many more words correctly compared with e-speak.  And I use it very well with the NVDA speech dictionary.  Are you familiar with the SAPI 5 demo you can try of Eloquence?  Eloquence and Via Voice, the same synthesizer from a sound and performance standpoint, are the most accurate in terms of proper pronounciation of words in my experience. 
 
Gene
----- Original Message -----
From: nvda@nvda.groups.io <nvda@nvda.groups.io> On Behalf Of Sharni-Lee Ward
Sent: Saturday, March 24, 2018 10:27 PM
To: nvda@nvda.groups.io
Subject: [nvda] I take it there have been no improvements to the flawed espeak ng in 2017.4?

I've just caught up on all my emails. If I understood an earlier thread correctly, despite our complaints when 2017.4 came out, the devs did not roll back the version of espeak packaged with NVDA to the version in 2017.3, which means that I will not be updating NVDA, despite the wonderful-sounding new features.


Espeak, and subsequently Espeak NG, has been my default synth since I started using NVDA. There was an issue in 2014 but I was able to get used to that, and then it got fixed, so I was contented. The problems I found in the espeak packaged with 2017.4 were more than I could stand, as someone who primarily uses NVDA to read fanfiction and my own stories. However, Espeak is among the only synths compatible with NVDA's inflection slider, which I have maxed out for optimal performance. I'm not about to sacrifice broad inflection for the drones of MS voices, and I'm currently using a Win 7 computer, but even if I was on a Win 10 one, the 1Core voices don't do it for me either. They're pleasant, but also flat. No inflection. No way to change the inflection. I can't stand human-sounding voices because they're less flexible than synthetic ones in this regard. Also, nearly all non Espeak voices are slow as well as flat, partially due to the length of pauses between clauses of a sentence, and this is about as unbearable as the flatness. And there is one more thing that really clinches my unwillingness to change synths. I cannot get anything else to cooperate with my speech dictionary. As I said, I read a lot of fanfiction, and I have put lots of characters'
names and the names of Pokémon into my dictionary so that they are pronounced properly. The other synths simply do not like these names, no matter what I try to do with them, and some of them even pronounced my name wrong, so I gave up on them.


This is why I want whoever's developing Espeak to roll back their changes to a previous version (surely they have older versions backed up somewhere. I would if I was developing something like this), and for NVDA to likewise roll back the version packaged with it until the devs of the synth work out how to alter one language without screwing up twenty others in the process. I know I can only go so far before I am forced to update to the latest version of NVDA or be left to languish as more and more things become incompatible with the last tolerable version I have.


Please, I am begging you, do something about this.









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