[orca-list] gnu-speach?

[Tony Baechler]

Similarly to the below post from the Orca list, is there any chance that a
connector for Speakup could be written to use this? The web page looks
interesting. I'm off to compile it now, if nothing else to see if the
speech quality is as good as the web page says. I would like other opinions
on this if anyone here has actually used it.


-------- Forwarded Message --------
Subject: [orca-list] gnu-speach?
Date: Thu, 22 Oct 2015 10:33:43 +0100
From: Majid Hussain <***@gmail.com>
To: orca-***@gnome.org

hi there,
this sounded interesting.
http://www.gnu.org/software/gnuspeech/
since it can create speach, could something be implimented with orca?
or am i just talking trees?
Majid HUssain
_______________________________________________
orca-list mailing list
orca-***@gnome.org
https://mail.gnome.org/mailman/listinfo/orca-list
Orca wiki: https://wiki.gnome.org/Projects/Orca
Orca documentation: https://help.gnome.org/users/orca/stable/
GNOME Universal Access guide:
https://help.gnome.org/users/gnome-help/stable/a11y.html
Log bugs and feature requests at http://bugzilla.gnome.org

[Jason White]

Post by Tony Baechler
Similarly to the below post from the Orca list, is there any chance that a
connector for Speakup could be written to use this? The web page looks
interesting. I'm off to compile it now, if nothing else to see if the
speech quality is as good as the web page says. I would like other opinions
on this if anyone here has actually used it.

I compiled and tested it at one point. I can't remember how many years ago
this was done. The speech was difficult to understand and not of high quality.
However, it may have improved since then. The original developers were speech
synthesis researchers and the software had to be ported to Linux from a code
base originally written for NextStep.

A promising project of more recent origin is OpenMary, which, although written
i Java, seems to offer clear and readily understandable speech, judging by the
online demonstration site. (The demonstration defaults to German; be sure to
change the language if you're testing in English.)

[Willem van der Walt]

Hi,
I grabbed the demo .wav and am not impressed.
Our own tts, speect, is better.
It is available at http://speect.sf.net
It is using hmm tts methods.
It is more natural sounding than gnu speech or espeak, and not IMHO suited
for screenreader use.
HMM is synthesized sound, modeled after human voices which is actual
recordings, IE. the computer is tought how speech should sound by giving
it examples of how a real person would read some given text.
The commercial version of our system, called qfrency, can be listened to
at http://www.qfrency.com.
Speect sounds the same.
FYI, Willem

Post by Jason White

Post by Tony Baechler
Similarly to the below post from the Orca list, is there any chance that a
connector for Speakup could be written to use this? The web page looks
interesting. I'm off to compile it now, if nothing else to see if the
speech quality is as good as the web page says. I would like other opinions
on this if anyone here has actually used it.

I compiled and tested it at one point. I can't remember how many years ago
this was done. The speech was difficult to understand and not of high quality.
However, it may have improved since then. The original developers were speech
synthesis researchers and the software had to be ported to Linux from a code
base originally written for NextStep.
A promising project of more recent origin is OpenMary, which, although written
i Java, seems to offer clear and readily understandable speech, judging by the
online demonstration site. (The demonstration defaults to German; be sure to
change the language if you're testing in English.)
_______________________________________________
Speakup mailing list
http://linux-speakup.org/cgi-bin/mailman/listinfo/speakup
--
This message is subject to the CSIR's copyright terms and conditions, e-mail legal notice, and implemented Open Document Format (ODF) standard.
The full disclaimer details can be found at http://www.csir.co.za/disclaimer.html.
This message has been scanned for viruses and dangerous content by MailScanner,
and is believed to be clean.
Please consider the environment before printing this email.

[Tony Baechler]

Post by Willem van der Walt
I grabbed the demo .wav and am not impressed.
Our own tts, speect, is better.
It is available at http://speect.sf.net
It is using hmm tts methods.
It is more natural sounding than gnu speech or espeak, and not IMHO suited
for screenreader use.
HMM is synthesized sound, modeled after human voices which is actual
recordings, IE. the computer is tought how speech should sound by giving it
examples of how a real person would read some given text.

I looked at the download site with interest. First, it would appear that
fully installed, it's bigger than Gnuspeech. Gnuspeech without a dictionary
weighs in at under 1 MB and with the data files at 3.8 MB. However, I
wasn't able to get the Ubuntu package to install on 14.04.3, even after
adding the PPA repository. Apparently packages are only available for 12.04
Precise. That's too bad as I am always willing to give other synthesizers a
chance. I'm always looking for high quality speech for reading text files,
ebooks, etc. If the packages get updated, please post here.

Also, to keep this on topic for the Speakup list, what about writing either
a module for Speech Dispatcher or a Speakup connector? If it's as good as
you claim it is, Orca users would probably also be interested, but there
would need to be a module for Speech Dispatcher.

[Willem van der Walt]

Hi,
One can compile from source and then also the hts engine wich is mensioned
in the readme.
I think if there are packages, they are likely outdated.
I cannot remember if there are sample voices on the site, if not, contact
me when you have built it.
There are both python and C apis, so eventually writing something like
espeakup, but for speect, should be relatively easy.
one can add -dwith_tests or something to the cmake command and will then
get a binary synth_test that can generate a .wav file. Putting that in a
pipe using play from sox can
be used as a basic test binary and could be used with speech-dispatcher's
generic module.
Note with speect as it is, you might have to do your own text
normalization.
Kind regards, Willem

Post by Tony Baechler

Post by Willem van der Walt
I grabbed the demo .wav and am not impressed.
Our own tts, speect, is better.
It is available at http://speect.sf.net
It is using hmm tts methods.
It is more natural sounding than gnu speech or espeak, and not IMHO suited
for screenreader use.
HMM is synthesized sound, modeled after human voices which is actual
recordings, IE. the computer is tought how speech should sound by giving it
examples of how a real person would read some given text.

I looked at the download site with interest. First, it would appear that
fully installed, it's bigger than Gnuspeech. Gnuspeech without a dictionary
weighs in at under 1 MB and with the data files at 3.8 MB. However, I
wasn't able to get the Ubuntu package to install on 14.04.3, even after
adding the PPA repository. Apparently packages are only available for 12.04
Precise. That's too bad as I am always willing to give other synthesizers a
chance. I'm always looking for high quality speech for reading text files,
ebooks, etc. If the packages get updated, please post here.
Also, to keep this on topic for the Speakup list, what about writing either
a module for Speech Dispatcher or a Speakup connector? If it's as good as
you claim it is, Orca users would probably also be interested, but there
would need to be a module for Speech Dispatcher.
_______________________________________________
Speakup mailing list
http://linux-speakup.org/cgi-bin/mailman/listinfo/speakup
--
This message is subject to the CSIR's copyright terms and conditions, e-mail
legal notice, and implemented Open Document Format (ODF) standard.
The full disclaimer details can be found at
http://www.csir.co.za/disclaimer.html.
This message has been scanned for viruses and dangerous content by MailScanner,
and is believed to be clean.
Please consider the environment before printing this email.