Date   

[espeak-ng:master] reported: espeak does not connect gen case in Farsi / Persian with E #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #862 espeak does not connect gen case in Farsi / Persian with E
By HKalbasi:

Yes it is possible but a built-in solution inside espeak-ng would be way more useful. Very tools, like screen readers have out of box support for espeak-ng but it is not easy to use hazm in them in this way.


[espeak-ng:master] reported: espeak does not connect gen case in Farsi / Persian with E #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #862 espeak does not connect gen case in Farsi / Persian with E
By jaacoppi:

Is it possible to input the text to hazm and get the correct text as output, and then pipe that to espeak-ng? If yes, you could try a script that pipes everything through hazm first. Something like hazm "text" | espeak-ng -v fa

I don't know if combining hazm and espeak-ng is simple because hazm is a python script.


[espeak-ng:master] closed: pronunciation error with abbreviations followed by "'s"#github

espeak-ng@groups.io Integration <espeak-ng@...>
 


[espeak-ng:master] reported: Position parameter of espeak_Synth doesn't work properly #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #861 Position parameter of espeak_Synth doesn't work properly
By jaacoppi:

Looks like POS_CHARACTER makes the speech start from the next full word after position X. See this:

#include <string.h>
#include <espeak-ng/espeak_ng.h>

static void say(char *text, int index) {
    espeak_Synth(text, strlen(text) +1, index, POS_CHARACTER, 0, espeakCHARS_AUTO,
                 NULL, NULL);
    espeak_Synchronize();
}
int main() {
    espeak_Initialize(AUDIO_OUTPUT_PLAYBACK, 0, NULL, 0);
    say("This starts from the 2nd word - starts.", 4);
    say("This starts from the 3rd word - from.", 8);
    say("This starts from the 4rd word - the.", 13);

}


The reason is here: https://github.com/espeak-ng/espeak-ng/blob/9e9c0a042b23d1b0ecd9871b5b60d56e9aeeca2f/src/libespeak-ng/readclause.c#L676-L683

It's clearly intended behaviour, but is it a good idea? I think not, it should break the word and start from the character position indicated.


[espeak-ng:master] reported: pronunciation error with abbreviations followed by "'s" #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #706 pronunciation error with abbreviations followed by "'s"
By jaacoppi:

Fixed in master, please close this ticket.


[espeak-ng:master] reported: Fixed announcer variant. #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Pull Request #863 Fixed announcer variant.
By 0x5066:

Interesting to know for the future


Pull Request Opened #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng/espeak-ng] Pull request opened by datajake1999:

#863 Fixed announcer variant.

It wasn't showing up because their were spaces in the name, so I removed the spaces and it now shows up.


[espeak-ng:master] reported: Some wrong pronunciations of German words #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #809 Some wrong pronunciations of German words
By JacekRuzyczka:

May I submit you another bunch of mispronounciations of German words containing the sp and st digraphs, which are erroneously pronounced as /sp/ and /st/ instead of /ʃp/ and /ʃt/, respectively? Faulty digraphs are marked bold:

B3 Northeim-Nord: für beide Richtungen nur ein Fahrstreifen abwechselnd frei. L382 Langenforther Straße Kugelfangtrift Sutelstraße: geänderte Verkehrsführung. L116 Lamstedt: Baustelle.


[espeak-ng:master] new issue: espeak does not connect gen case in Farsi / Persian with E #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Issue Created by HKalbasi:
#862 espeak does not connect gen case in Farsi / Persian with E

This is a huge problem with espeak Farsi. For example: Physic laws are beautiful.

قوانین فیزیک جالب هستند

espeak say it:

ghavaanin fizik jaleb hastand

but the correct is:

ghavaanine fizik jaleb hastand

notice the "E". This problem is very common, I can say in every sentence we have this E connections and this means espeak reads every sentence wrong :)

hazm library can detect this but it is based on machine learning and needs a 20 megabytes model. Can we use this in espeak-ng? For example by making Farsi a seperate installable package?


Updates to Github #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Issue Created by dieggsy:
#861 Position parameter of `espeak_Synth` doesn't work properly

As a minimal example:

#include <string.h>
#include <espeak-ng/espeak_ng.h>

static void say(char *text) {
    espeak_Initialize(AUDIO_OUTPUT_PLAYBACK, 0, NULL, 0);
    espeak_Synth(text, strlen(text) +1, 1, POS_CHARACTER, 0, espeakCHARS_AUTO,
                 NULL, NULL);
    espeak_Synchronize();
}
int main() {
    say("hello there");

}

Here, the position parameter being passed is 1, with position_type set as POS_CHARACTER. Thus I'd expect to hear "ello there", but instead it behaves exactly like POS_WORD.

With POS_WORD and POS_SENTENCE, the number seems to be 1-indexed, as it only skips the first word/sentence if the position parameter is set to 2. This doesn't appear to be documented anywhere.


[espeak-ng:master] New Issue Created by dieggsy:
#861 Position parameter of espeak_Synth doesn't work properly

As a minimal example:

#include <string.h>
#include <espeak-ng/espeak_ng.h>

static void say(char *text) {
    espeak_Initialize(AUDIO_OUTPUT_PLAYBACK, 0, NULL, 0);
    espeak_Synth(text, strlen(text) +1, 1, POS_CHARACTER, 0, espeakCHARS_AUTO,
                 NULL, NULL);
    espeak_Synchronize();
}
int main() {
    say("hello there");

}

Here, the position parameter being passed is 1, with position_type set as POS_CHARACTER. Thus I'd expect to hear "ello there", but instead it behaves exactly like POS_WORD.

With POS_WORD and POS_SENTENCE, the number seems to be 1-indexed, as it only skips the first word/sentence if the position parameter is set to 2. This doesn't appear to be documented anywhere.


Pull Request Opened #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng/espeak-ng] Pull request opened by ArenaL5:

#860 Small update for F-Droid

I'm sorry I couldn't update this sooner. This is a continuation of PR #820 to list eSpeak NG in F-Droid.

They recommend having short descriptions shorter than 80 characters to prevent F-Droid clients from truncating them, so I'm just doing that with this new commit.

They also recommend increasing versionCode in every release, to allow auto-updates. In particular, they ask that the current versionCode (20) is larger than the versionCode for the latest tag release.


Updates to Github #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Issue Created by sthibaul:
#859 Replacement character (U+FFFD) completely disturbs speak

Hello,

With espeak-ng 1.50, for instance when running

echo "hého hého" | espeak-ng -v fr --stdin

I get the same word spoken twice (with the 'é' properly spoken). But when running

echo "hého � hého" | espeak-ng -v fr --stdin

i.e. with a replacement character (U+FFFD) between the two words, the second word gets completely mangled, it looks like espeak-ng is suddently changing the way it decodes non-ascii characters

Such replacement character is a relatively common thing to encounter, and when properly encoded in utf-8 as exampled above, is not a sign that character decoding should change.


[espeak-ng:master] New Issue Created by sthibaul:
#859 Replacement character (U+FFFD) completely disturbs speak

Hello,

With espeak-ng 1.50, for instance when running

echo "hého hého" | espeak-ng -v fr --stdin

I get the same word spoken twice (with the 'é' properly spoken). But when running

echo "hého � hého" | espeak-ng -v fr --stdin

i.e. with a replacement character (U+FFFD) between the two words, the second word gets completely mangled, it looks like espeak-ng is suddenly changing the way it decodes non-ascii characters

Such replacement characterhttps://github.com/espeak-ng/espeak-ng/issues/859 is a relatively common thing to encounter, and when properly encoded in utf-8 as exampled above, is not a sign that character decoding should change.


[espeak-ng:master] New Issue Created by sthibaul:
#859 Replacement character (U+FFFD) completely disturbs speak

Hello,

With espeak-ng 1.50, for instance when running

echo "hého hého" | espeak-ng -v fr --stdin

I get the same word spoken twice (with the 'é' properly spoken). But when running

echo "hého � hého" | espeak-ng -v fr --stdin

i.e. with a replacement character (U+FFFD) between the two words, the second word gets completely mangled, it looks like espeak-ng is suddenly changing the way it decodes non-ascii characters

Such replacement character is a relatively common thing to encounter, and when properly encoded in utf-8 as exampled above, is not a sign that character decoding should change.


Pull Request Updated #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng/espeak-ng] Pull request updated by demaere-oiie:

#858 qdb - lang belta (conlang from The Expanse)

This is a TTS package for lang belta, an artificial creole developed by Nick Farmer for "The Expanse" TV show. (NB: it differs significantly from the "belter creole" language in the books) It is an unofficial fan creation.

At the moment, it contains my best interpretation based on conversations with the speech coach for the show, Eric Armstrong, and people such as https://www.youtube.com/watch?v=wcgaIqEioH4 who have learned the language orally, directly from Farmer.

For phonology and references, see documentation at the bottom of the following page: https://glaemscrafu.jrrvf.com/english/glaemscribe.html?mode=lang_belta-tengwar-dadef

pedantic notes:

  • the necessary phonemes were cut-and-pasted from other espeak-ng languages, en-us, en-rp, and jp.
  • there may be more aspects of the language revealed in upcoming seasons
  • we only have numbers up to (at most) 9'999 at this time
  • the 'qdb' abbreviation is q-private and has not been asssigned. 'db' is for da belte, the language's toponym


Pull Request Updated #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng/espeak-ng] Pull request updated by demaere-oiie:

#858 qdb - lang belta (conlang from The Expanse)

This is a TTS package for lang belta, an artificial creole developed by Nick Farmer for "The Expanse" TV show. (NB: it differs significantly from the "belter creole" language in the books) It is an unofficial fan creation.

At the moment, it contains my best interpretation based on conversations with the speech coach for the show, Eric Armstrong, and people such as https://www.youtube.com/watch?v=wcgaIqEioH4 who have learned the language orally, directly from Farmer.

For phonology and references, see documentation at the bottom of the following page: https://glaemscrafu.jrrvf.com/english/glaemscribe.html?mode=lang_belta-tengwar-dadef

pedantic notes:

  • the necessary phonemes were cut-and-pasted from other espeak-ng languages, en-us, en-rp, and jp.
  • there may be more aspects of the language revealed in upcoming seasons
  • we only have numbers up to (at most) 9'999 at this time
  • the 'qdb' abbreviation is q-private and has not been asssigned. 'db' is for da belte, the language's toponym


Pull Request Updated #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng/espeak-ng] Pull request updated by demaere-oiie:

#858 qdb - lang belta (conlang from The Expanse)

This is a TTS package for lang belta, an artificial creole developed by Nick Farmer for "The Expanse" TV show. (NB: it differs significantly from the "belter creole" language in the books) It is an unofficial fan creation.

At the moment, it contains my best interpretation based on conversations with the speech coach for the show, Eric Armstrong, and people such as https://www.youtube.com/watch?v=wcgaIqEioH4 who have learned the language orally, directly from Farmer.

For phonology and references, see documentation at the bottom of the following page: https://glaemscrafu.jrrvf.com/english/glaemscribe.html?mode=lang_belta-tengwar-dadef

pedantic notes:

  • the necessary phonemes were cut-and-pasted from other espeak-ng languages, en-us, en-rp, and jp.
  • there may be more aspects of the language revealed in upcoming seasons
  • we only have numbers up to (at most) 9'999 at this time
  • the 'qdb' abbreviation is q-private and has not been asssigned. 'db' is for da belte, the language's toponym


Pull Request Updated #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng/espeak-ng] Pull request updated by demaere-oiie:

#858 qdb - lang belta (conlang from The Expanse)

This is a TTS package for lang belta, an artificial creole developed by Nick Farmer for "The Expanse" TV show. (NB: it differs significantly from the "belter creole" language in the books) It is an unofficial fan creation.

At the moment, it contains my best interpretation based on conversations with the speech coach for the show, Eric Armstrong, and people such as https://www.youtube.com/watch?v=wcgaIqEioH4 who have learned the language orally, directly from Farmer.

For phonology and references, see documentation at the bottom of the following page: https://glaemscrafu.jrrvf.com/english/glaemscribe.html?mode=lang_belta-tengwar-dadef

pedantic notes:

  • the necessary phonemes were cut-and-pasted from other espeak-ng languages, en-us, en-rp, and jp.
  • there may be more aspects of the language revealed in upcoming seasons
  • we only have numbers up to (at most) 9'999 at this time
  • the 'qdb' abbreviation is q-private and has not been asssigned. 'db' is for da belte, the language's toponym


Pull Request Opened #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng/espeak-ng] Pull request opened by demaere-oiie:

#858 qdb - lang belta (conlang from The Expanse)

This is a TTS package for lang belta, an artificial creole developed by Nick Farmer for "The Expanse" TV show. (NB: it differs significantly from the "belter creole" language in the books) It is an unofficial fan creation.

At the moment, it contains my best interpretation based on conversations with the speech coach for the show, Eric Armstrong, and people such as https://www.youtube.com/watch?v=wcgaIqEioH4 who have learned the language orally, directly from Farmer.

For phonology and references, see documentation at the bottom of the following page: https://glaemscrafu.jrrvf.com/english/glaemscribe.html?mode=lang_belta-tengwar-dadef

pedantic notes:

  • the necessary phonemes were cut-and-pasted from other espeak-ng languages, en-us, en-rp, and jp.
  • there may be more aspects of the language revealed in upcoming seasons
  • we only have numbers up to (at most) 9'999 at this time
  • the 'qdb' abbreviation is q-private and has not been asssigned. 'db' is for da belte, the language's toponym


[espeak-ng:master] new issue: New release of espeak #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Issue Created by feerrenrut:
#857 New release of espeak

As per: #792

Over at nvaccess/nvda we have started our 2021.1 release process by updating the version of espeak-ng included. We selected the latest espeak-ng commit (at the time) and would love it if an espeak-ng release process could be started that we could use for the final release.

We have updated to espeak-ng commit 82d5b7b04488412845101851f36da6953cac4378F The NVDA 2021.1 release will be made in the first quarter of 2021.


[espeak-ng:master] new issue: Cross compiling not work #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Issue Created by leobbb:
#856 Cross compiling not work

Hi, I want to build espeak-ng dynamic library for RK platform, but it seems not work. build: x86_64-linux-gnu, host: aarch64-linux-gnu

I use command
CC=gcc ./configure --host=aarch64-linux-gnu make -B src/espeak-ng src/speak-ng

Build system had done and no error was reported. But, i found the type of espeak-ng library not changed. readelf -h src/.libs/libespeak-ng.so.1.1.51 ELF Header: Magic: 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 Class: ELF64 Data: 2's complement, little endian Version: 1 (current) OS/ABI: UNIX - System V ABI Version: 0 Type: DYN (Shared object file) Machine: Advanced Micro Devices X86-64

Please tell me how to make it work? Thank you for your help.


[espeak-ng:master] new issue: Greek: Pronunciation corrections #github

espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Issue Created by Zseli22:
#855 Greek: Pronunciation corrections

My native language is Hungarian, and I have been using Espeak for many years due to it being the default speech engine of NVDA. However, I am learning greek as a second language and have noticed some oddities in the translations done by the engine. I have looked at the rule files and determined that the last change took place before the move over to Espeak NG, meaning the language has not been updated in over three years. I am not familiar with the process of cloning and testing new additions to the rules file, but I would like to propose some changes so other contributers will be able to understand what issues there are.

The thing that stands out to me immediately is the phonemes that are in use. Specifically the "t" sound and the "r" sound does not sound accurate. A simpler sound similar to other languages such as croatian or hungarian would be more accurate, in my opinion. The second section has to do with pronouncing numbers. This section is much easier to explain since the rules file gives exact pronunciations for the numbers, and they can easily be changed. If any part of this issue is to be tackled, this part would be the easier one. The number 5 pronounces the t when normally a "ντ" is supposed to result in a "nd" like sound instead. Curiously the numbers 50 or 500 get pronounced correctly, although every subsiquent tens didget calls the routine for the initial number 5 which gets pronounced incorrectly. The number 11 pronounces the "Δ" when in this case it's written as "ντ" and so must be pronounced with an "ND" like sound. The numbers 70 and 80 are not using "Γ" and "Δ" where applicable. E.g."Εβδομίντα" comes out as "εμπντομίντα". Finally, the phonemes called when anything in the hundreds series above 200 is said sounds extremely out of place. A normal "s" sound could replace it.

601 - 620 of 4292