New Espeak voices #4400

nvaccessAuto opened this Issue Aug 19, 2014 · 9 comments

1 participant


Reported by briang1 on 2014-08-19 07:22
It has been mentioned to me more than once that the Quincy voice is too quiet on some machines, and so I have done a new version which is considerably louder at normal volume settings.
Care needs to be taken of course not to overload the audio systems in laptops or rattle the speakers.
Blocking #5061


Attachment quincy2 added by briang1 on 2014-08-19 07:24
New Variant. Quincy 2 louder version of Quincy


Attachment quincy2.2 added by ateu on 2014-08-19 13:44
Changed just the internal name which was quincy, and woud cause conflict with the original quincy


Attachment added by briang1 on 2014-08-19 14:01
A zip file with other voices inside


Comment 1 by briang1 on 2014-08-19 14:03
Im not suggesting we should add them all, depends on what people think, just wanted to shove them

The voices in the file are:
Belinda - Sounds best with English north or west mids accents

Dennis_2 is a louder version of Dennis

Lee2 is a louder version of Lee. I'm told this is a good voice for some languages, but to me its L sounds like R

Max2 is a louder version of Max

Robert_2 is a louder version of robert, with improved clarity to my ears.

Steph is a voice that sounds reasonably good without the peakiness of other female voices, but is a little quiet.
Steph2 and Steph 3 are slightly different versions both a little louder, but with slightly more bite.

Susan, Susan 3 and Susan 5 are dry, ie no echo, versions of some of the standard female variants. They are louder, but a bit strident.

I added the above but seem to have lost it in the editing last time
in a place where they were safe


Comment 2 by blindbhavya on 2014-08-19 14:48
Is Steph 2 the same as Steph_test 1 shared by another member?
If not please could you attach it here?
If yes then James Sir please listen to that variant first because it sounds very pleasant.


Comment 3 by briang1 on 2014-08-20 08:21
I have no idea as I've not seen that one. Steph was around for a while, but did not get into an nvda release I think. The one in the zip file is probably almost the same as that one. the other two Stephs are basically the same but with different voicing and other values so they appear louder, but can also tend to get into the same issues as the other female varients.
As has been said before, the problem for some people in espeak is not the actual voice, but the phonemes used by espeak. Some describe it as too British!


Comment 5 by blindbhavya on 2014-10-27 06:09
Actually, these days, I have tested in detail, the Steph 3 voice that you had shared on an NVDA list some months ago and compared it carefully with Steph_test 1
To me, the Steph 3 variant is certainly a bit louder than Steph_test 1, and much louder than Steph.
Though I can understand and use all other variants, I use Steph 3 as my default one.
I personally haven't really tested Edward, but till what I have heard it, I definitely prefer Steph 3 over it.
All these are my views, and if possible, I would request others to try it as well.


Comment 6 by briang1 on 2014-10-27 10:45
In getting Steph to be louder, I had to make some compromises which did result in it being less smooth. I imagine these are limitations of the way these sort of voices are built in the first place.
There are limits to what one can do using the standard built in phonemes I suggest.

That is why the Klat and hence the updated speech player does sound quite different.

Everyone is different though, and I know some people cannot stand the current default while others feel its good. I now use Quincy 2, but then I did tweak the Quincy varients to my own ears.


Comment 8 by JamaicanUser (in reply to comment 7) on 2015-04-27 13:50
is at least adding these voices to the next release an issue?

Replying to jteh:

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment