Date   

Updates to Github #github

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

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

#791 add output hooks for wavegen

I've added some hooks to wavegen, which can be used to output aligned speech. This information can be used by prosody transplantation tools similar to MBROLIGN. There is also a function to set a constant f0 in Hz, which may be used by some prosody transplantation tools. If none of the functions is called, the behaviour of espeak-ng is unchanged.


[espeak-ng:master] New Comment on Pull Request #790 improvement japanese
By tplaten:

this looks good for me


Pull Request Opened #github

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

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

#790 improvement japanese

Hello


[espeak-ng:master] reported: Arch Linux aarch64 on Android and speak char delay issue #github

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

[espeak-ng:master] New Comment on Issue #789 Arch Linux aarch64 on Android and speak char delay issue
By valdisvi:

espeak-ng source code can be compiled with different options. Maybe it helps, if it is compiled with --with-async=no and --with-sonic=no or --with-pcaudiolib=no (or some other more technical/specific) settings.


[espeak-ng:master] new issue: Arch Linux aarch64 on Android and speak char delay issue #github

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

[espeak-ng:master] New Issue Created by JanuszChmiel:
#789 Arch Linux aarch64 on Android and speak char delay issue

Dear specialists, I Am using latest stable ArchLinux aarch64 which run on A top of Android Termux and Proot. I Am using Orca screen reader with Speech-dispatcher 0.9.1. I use Espeak-ng without pulseaudio-alsa module. Espeak-ng work perfectly except The situation, when Orca uses char by char movement by using arrow keys for example. In this case, Espeak-ng skip The char and speak second one or speak some characters twice. Termux uses patched Pulseaudio which uses Opensless protocol to communicate with other Android subsystems which are responsible for sound output. As A result, Pulseaudio is stable. But to be able to use this sound server from running ArchLinux aarch64 I muse use echo "export PULSE_SERVER=\"127.0.0.1\"" >> ~/.bashrc I Am running Pulseaudio from Termux /etc/profile by using those commands.

pulseaudio --start --exit-idle-time=-1 pacmd load-module module-native-protocol-tcp auth-ip-acl=127.0.0.1:4712 auth-anonymous=1 Sure, i must also have pulseaudio installed on my Arch Linux, because some libraryes are needed to transfer sound data to Pulseaudio sound server which run in Termux.

By other words. When I press arrow keys at The time, when previously spoken character is still being pronounced by Espeak-ng response is good and no skipping occurs. But when I press arrow key in The moment, when Espeak-ng do not pronounce previous character, there is skipping problem.

In some cases, when I type characters on The keyboard, some characters are also skipped.

My working solution have been to combina pulseaudio-alsa ArchLinux package in combination with old Espeak. In this case, there is newer a character skipping. But unfortunately, Speech-dispatcher even latest development Github code simply refuses to cooperate with pulseaudio-alsa when Espeak-ng is being used. 0.9.1 version support pulseaudio-alsa and old Espeak for reliable speech support. But pulseaudio-alsa can not be used in conjunction with Espeak-ng. In Speech-dispatcher 0.9.1 and also in The newest Speech-dispatcher development code.

I Am sad, because using still old espea which will not be developed is also not A solution. Some ArchLinux repository manager can remove pulseaudio-alsa sometimes and what could I do in this case.

How to precisely help you to debug Espeak-ng skipping?

Running Pulseaudio by using simple protocol from Arc Linux and having Pulseaudio for Termux disabled is not good idea. Because two available Simple protocol players for ANdroid do not implement special feature of Speech-dispatcher, which makes sound chip online ready. Transistors are producing soft sound and as A result, The delay between key press and getting The feetback rom Espeak-ng is fast but only if Pulseaudio for Termux is being used. Not if I configure Pulseaudio directly in Arch Linux and if Pulseaudio will only stream The sound by using Simple protocol. The next problem is memory allocation based. It is necessary to softly tune The buffer lenght in The speech-dispatcher coniguration file and also in Simple protocol player. In other case, Pulseaudio sound server will crash randomly thanks to memory allocation issues.

Thank you very much for yours help.


Updates to Github #github

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

2 New Commits:

[espeak-ng:master] By Icenowy Zheng <icenowy@...>:
00fe15cc28cc: cmn: use voiceless uvular fricative as pronounciation of pinyin h

According to [1] and [2] , the pronounciation of the pinyin character
h is not accurately [x].
According to [2], [χ] is instead a possible pronounication in young
users.

As a young native speaker of Mandarin, I feel the sound of [χ] in espeak
more satisfied than [x] here.

[1] https://en.wikipedia.org/wiki/Standard_Chinese_phonology
[2] https://phesoca.com/aws/268/

Signed-off-by: Icenowy Zheng <icenowy@...>

Modified: dictsource/zh_rules


[espeak-ng:master] By Valdis Vitolins <valdis.vitolins@...>:
22a30434038d: Merge pull request #788

Modified: dictsource/zh_rules


[espeak-ng/espeak-ng] Pull request closed by valdisvi:

#788 cmn: use voiceless uvular fricative as pronounciation of pinyin h

According to [1] and [2] , the pronounciation of the pinyin character h is not accurately [x]. According to [2], [χ] is instead a possible pronounication in young users.

As a young native speaker of Mandarin, I feel the sound of [χ] in espeak more satisfied than [x] here.

[1] https://en.wikipedia.org/wiki/Standard_Chinese_phonology [2] https://phesoca.com/aws/268/

Signed-off-by: Icenowy Zheng icenowy@...


[espeak-ng:master] reported: cmn: use voiceless uvular fricative as pronounciation of pinyin h #github

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

[espeak-ng:master] New Comment on Pull Request #788 cmn: use voiceless uvular fricative as pronounciation of pinyin h
By Icenowy:

BTW as /x/ is still a commonly accepted mark of this phoneme (although even /h/ is an acceptable variant), should I override its IPA output to /x/ ?


Pull Request Opened #github

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

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

#788 cmn: use voiceless uvular fricative as pronounciation of pinyin h

According to [1] and [2] , the pronounciation of the pinyin character h is not accurately [x]. According to [2], [χ] is instead a possible pronounication in young users.

As a young native speaker of Mandarin, I feel the sound of [χ] in espeak more satisfied than [x] here.

[1] https://en.wikipedia.org/wiki/Standard_Chinese_phonology [2] https://phesoca.com/aws/268/

Signed-off-by: Icenowy Zheng icenowy@...


Updates to Github #github

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

[espeak-ng:master] New Comment on Pull Request #787 fix 492
By valdisvi:

If you use project on odo.lv, then just choose task pull from upstream.


[espeak-ng:master] New Comment on Pull Request #787 fix 492
By hozosch:

Ah yes, you meant that. Yes, already pushed to my project.


Updates to Github #github

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

[espeak-ng:master] New Comment on Pull Request #787 fix 492
By valdisvi:

@hozosch, I merged this pull request with slight change — I changed your comment fix 492 to Fix issue #492 (note hash before number). This change may require you to overwrite history of local repository with history from the main espeak-ng repository. When mention issues, please use that format, because in that way GitHub automatically generates link to the issue. Because of your question about build failure I noticed, that Travis builds were failing because of incorrect MBROLA test, which I fixed.


[espeak-ng/espeak-ng] Pull request closed by valdisvi:

#787 fix 492


[espeak-ng:master] New Comment on Pull Request #787 fix 492
By valdisvi:

@hozosch, I merged this pull request with slight change — I changed your comment fix 492 to Fix issue #492 (note hash before number). This change may require you to overwrite history of local repository with history from the main espeak-ng repository. When mention issues, please use that format, because in that way GitHub automatically generates link to the issue. Because of your question about build failure I noticed, that Travis builds were failing because of incorrect MBROLA test, which I fixed.


[espeak-ng:master] New Comment on Issue #778 Cantonese have some numbers missing when reading
By valdisvi:

That part of code can be made easier to read (e.g. block of if (name2 == L3('z', 'h', 'y')) {..} can be moved to case L3('c', 'm', 'n'):...), but, I think, it works as expected (note that cases without break; are not exclusive). I found, that problem is somewhere deep into translating numbers. If I comment out other powers of digits in lv_list and leave just:

// numbers
_0      n'ulle
_1      v'iens
_2      d'ivi
_3      tRi:s
_4      tS'etRi
_5      p'ietsi
_6      s'eSi
_7      s'eptin^i
_8      'astuon^i
_9      d'evin^i
//... other lines are commented out

it spells in the same wrong way as for Mandarin:

espeak-ng -vlv -X "2019"
Found: '_2' [d'ivi]
Found: '_9' [d'evin^i]
d'ivi d'evin^i

And when I thought I know where problem my be hiding, I noted that it works for languages like K'iche' and Pyash, where also only single digits are defined (quc_list and py_list accordingly):

espeak-ng -vquc -X '2019'
Translate '2019'
Found: '_2' [keb']
Found: '_0' [majb'al]
Found: '_1' [jun]
Found: '_9' [b'elejeb']
kebm'ajb'al_!junb'elej,eb_!

espeak-ng -vpy -X '2019'
Translate '2019'
Found: '_2' [tjut]
Found: '_0' [zRon]
Found: '_1' [hjik]
Found: '_9' [twun]
tj'utzRon_!hj,iktwun_!


[espeak-ng:master] New Comment on Pull Request #787 fix 492
By hozosch:

How do I go about changing the history?


Github push to espeak-ng:espeak-ng #github

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

3 New Commits:

[espeak-ng:master] By Karl Eick <hozosch@...>:
830690be0d9c: Fix issue #492

Modified: phsource/ph_german


[espeak-ng:master] By Karl Eick <hozosch@...>:
1c76c01ee09a: Update hashsum

Modified: tests/language-phonemes.test


[espeak-ng:master] By Valdis Vitolins <valdis.vitolins@...>:
21683e0d0e3d: Merge pull request #787

Modified: phsource/ph_german
Modified: tests/language-phonemes.test


Updates to Github #github

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

[espeak-ng:master] New Comment on Pull Request #787 fix 492
By hozosch:

Well thanks for that info. Now I also see what I was after in my comment. What I meant is that I didn't see the results at all, no matter if they failed or succeded, so couldn't retrieve the hashsum which is what I was looking for. Now, it's there, and no, I don't mean your comment.


1 New Commit:

[espeak-ng:master] By Valdis Vitolins <valdis.vitolins@...>:
cef02afc59da: As MBROLA is optional, don't exit with error if mbrola is not found

Modified: tests/mbrola.test


[espeak-ng/espeak-ng] Pull request updated by hozosch:

#787 fix 492


[espeak-ng:master] reported: fix 492 #github

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

[espeak-ng:master] New Comment on Pull Request #787 fix 492
By valdisvi:

Because build is failing. When you change phoneme definitions you also have to update hashsum for phoneme tests in tests/langugage-phonemes.test file. In this particular case: from d05d865062b8d9d4b799358f2a912033b05cac28 to e20faa2bced0d5d6e56bc898dd12eaa75c3d2217


[espeak-ng:master] reported: fix 492 #github

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

[espeak-ng:master] New Comment on Pull Request #787 fix 492
By hozosch:

Why don't I see the travis build results?


Pull Request Opened #github

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

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

#787 fix 492


[espeak-ng:master] reported: phoneme definition to make usually voiced consonants voiceless if the next phoneme is a consonant #github

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

[espeak-ng:master] New Comment on Issue #492 phoneme definition to make usually voiced consonants voiceless if the next phoneme is a consonant
By valdisvi:

Rules in ph_... file are applied when text is already converted to the chain of phonemes to be synthesized, therefore, I think, "pause" in that stage is actually any silence between phonemes (sounds).


[espeak-ng:master] reported: phoneme definition to make usually voiced consonants voiceless if the next phoneme is a consonant #github

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

[espeak-ng:master] New Comment on Issue #492 phoneme definition to make usually voiced consonants voiceless if the next phoneme is a consonant
By hozosch:

Will this affect $brk as well?


[espeak-ng:master] reported: phoneme definition to make usually voiced consonants voiceless if the next phoneme is a consonant #github

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

[espeak-ng:master] New Comment on Issue #492 phoneme definition to make usually voiced consonants voiceless if the next phoneme is a consonant
By valdisvi:

Yes, there is option IF nextPh(isPause) ....


[espeak-ng:master] reported: phoneme definition to make usually voiced consonants voiceless if the next phoneme is a consonant #github

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

[espeak-ng:master] New Comment on Issue #492 phoneme definition to make usually voiced consonants voiceless if the next phoneme is a consonant
By hozosch:

This sounds more like something I can do, but what I mean is this: if there is a break defined, words ending in a voiced consonant are not unvoiced, like they should be. Can this be fixed somehow?


[espeak-ng:master] reported: phoneme definition to make usually voiced consonants voiceless if the next phoneme is a consonant #github

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

[espeak-ng:master] New Comment on Issue #492 phoneme definition to make usually voiced consonants voiceless if the next phoneme is a consonant
By valdisvi:

@hozosch, it will not look very clean and short, but you can make switch from voiced to voiceless consonant with rules in ph_german file. E.g. it is already done to to switch b to p if following is l2:

...
phoneme b
  vcd blb stp
  voicingswitch p
  IF nextPh(l2) THEN
    ChangePhoneme(p)
  ENDIF
...

You need to add similar condition statements for other phonemes and phonemes which may follow them.


Updates to Github #github

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

1 New Commit:

[espeak-ng:master] By Valdis Vitolins <valdis.vitolins@...>:
1fb68ffffea4: Fix issue #773: wrong pronunciation of Hawaiʻi

Modified: dictsource/en_rules
Modified: tests/translate.test


[espeak-ng:master] New Comment on Issue #773 Pronunciation of Hawaiʻi
By valdisvi:

As we don't know, how long it will take to fix it properly, for now I fixed with proposed workaround with replace rule, which works only for English.

961 - 980 of 4272