Hi,

I'm not sure if this is an issue with espeakup or perhaps espeak-ng-data-udeb
(which is at version 1.50+dfsg-7) -- please reassign as you see fit.

When installing Debian with debian-installer (using daily d-i images) in Speech Synthisis
mode, if one selects Russian the resulting audio seems garbled.

There is a recording here:

https://openqa.debian.net/tests/11385#step/locale/3

I'm not a Russian speaker myself, but having asked someone who is to listen to
this they report that they are unable to understand the audio even when the text
that is supposedly being read is in front of them.

Cheers, Phil.

Hello,

Philip Hands, le mer. 07 juil. 2021 12:25:02 +0200, a ecrit:

espeak-ng -v ru "Выберите местонахождение"

seems to be producing the same output, so reassigning to espeak-ng.

Hum, that's bad. Does it help if we slow the speed down? E.g.

espeak-ng -s 100 -v ru "Выберите местонахождение"

Possibly it's just the default speed which is bad for russian. Trying
translate.google.com, it seems that -s 100 provides the same kind of
speed.

Samuel

Samuel Thibault <sthibault@debian.org> writes:

I guess we need a native Russian speaker to have a listen.

The person I asked is fluent but not native and suspects that someone
who's familiar with the synthesised voice, and a native, might have a
better time with it.

She has the impression that the phonemes are all being given an equal
amount of time/emphasis when they should get varying emphasis (or
something like that) and that is making it very hard to understand.

I note that changing the speed settings in english seems to make rather
more difference to the outcome than is seems to in russian, but perhaps
that's just because I can follow what's going on in english.

Cheers, Phil.