Bug#591717: [Espeak-general] Bug#591717: espeak: Library unble to open soundcard.

2010-08-29 Thread Jonathan Duddington
On 22 Aug, Samuel Thibault samuel.thiba...@ens-lyon.org wrote:

 One of our debian users has troubles with using espeak on his machine,
 getting 

 wave_open_sound  Pa_OpenStream : err=-9996 (Invalid device)

This error message comes from espeak, from file wave.cpp at line 307.

This is in function wave_open_sound() which tries to open a sound
channel for output.  Error number -9996 means invalid device, but I
don't know what that means in this context.

espeak tries to open a sound channel for 22050 Hz, 16 bit, mono.

If the error is -9998 (Invalid channel count) it will try stereo, but
that is not the case here.

If 22050 Hz is not available, the error should have been -9997 (Invalid
bit rate).  Error Invalid device may suggest something wrong with
ALSA data.

eSpeak can be made to use a specified device number (I'm not sure what
device number means), by a line in the file  espeak-data/config

You could try adding a line:

pa_device  1

in the file espeak-data/config, and change 1 to a number that works.

Is this an index into a some ALSA data?  This number sets the value of
variable option_device_number which is used at line 503 in function
select_device() in wave.cpp.  I don't know much about this code.  It
was written by Gilles Casse gca...@oralux.org, so he may have a
better understanding of this problem.




-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#570476: espeak give a segmentation fault error message when it is ran by the user

2010-02-18 Thread Jonathan Duddington
On 19 Feb, Norberto Feliberty ngame0...@aol.com wrote:
 Package: espeak
 Version: 1.41.01-1

 When running espeak it gives the error message of segmentation fault.  
 Architecture: powerpc (ppc64)

Is this a big-endian processor?
This problem may be caused by little-endian data in the espeak-data
directory (/usr/shared/espeak-data).  This affects the files
espeak-data/phondata,phonindex,phontab.

You can make a big-endian version of these files a follows:

1.  Download espeak-1.41.01-source.zip from
http://sourceforge.net/projects/espeak

2.  In this download, find and enter the directory:
platforms/big-endian

3.   Do make to compile the espeak-phoneme-data program.

4.  Run the espeak-phoneme-data program binary.  This will make new
version of the files:  phondata, phonindex, phontab.

5.  Copy these 3 files into /usr/shared/espeak-data  to replace the
original files.




-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#557311: Facility to produce phonic letter sounds

2009-11-21 Thread Jonathan Duddington
On 21 Nov, markhob...@yahoo.co.uk wrote:

 I would like the facility to be able to produce the phonic sounds of 
 letters, rather than the names of the letters.

 If I type:

 espeak 'a, b, c, d, e, f, g'

 This gives ay bee see dee ee eff jee

 I am writing some educational software and I need to be able to
 produce the short sounds.

What exactly do you mean by phonic sounds of letters.

You can use phoneme codes to make sounds, eg:

espeak [['a]]

will say an a vowel, not the name of the letter a.

For most consonants, you need a vowel in order to produce the consonant
sound.  So for b try:

espeak [...@]]
or
espeak [...@-]]

You can fine what are the phoneme codes that eSpeak uses for a word, by
using the -x options, eg:

espeak -x test

shows the phonemes:
t'Est

I'm not sure what the phonic sounds for a,e,i,o,u will be, since each
vowel letter (and some consonants) have different sounds in different
words.  For British English the short vowels are phonemes:
[a] [E] [I] [0] [V], as in sat, set, hit, hot, hut.




-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#461072: Broken American (en-us) voices + PATCH

2008-01-22 Thread Jonathan Duddington
On 23 Jan, Alan Baghumian [EMAIL PROTECTED] wrote:

 I compared the 1.29 and 1.30 code and find that what differs. I
 rebuilt the pakage using the attached patch and us-mb3 voice is
 working again. Please add this patch to support atleast a single AmE
 accent, ready to use w/o any mambo jambo :-) 

 Patch to espeak-data/voices/mb/mb-us3
 -mbrola us3 us3_phtrans
 +mbrola us3

NO. This patch is wrong.
This difference is an error in eSpeak 1.29 which has been fixed in
eSpeak 1.30.  The new version with
  mbrola us3 us3_phtrans
is correct.

The eSpeak us-mbrola-3 voice does not produce speech.  Instead it
produces phoneme information (as text) which can be piped to mbrola so
that the mbrola voice produces speech.  See
http://espeak.sourceforge.net/mbrola.html

In eSpeak 1.29, the error meant that eSpeak could not produce mbrola
phoneme information, so it speaks the text using its own voice.

If you want to use eSpeak to speak by itself, without using mbrola,
then use:
  espeak -v en-r

en-r is eSpeak's American voice.  I haven't called it en-us
because it doesn't sound very American to me yet.  I think it's better
in the current version, eSpeak 1.31.  Perhaps native speakers can
suggest improvements.

To summarise, this is not an error, and this bug can be closed.
To achieve the result which you had using eSpeak 1.29, use:
  espeak -v en-r
not
  espeak -v us-mbrola-3




-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#461072: Broken American (en-us) voices

2008-01-18 Thread Jonathan Duddington
On 16 Jan, Alan Baghumian [EMAIL PROTECTED] wrote:

 It seems that non of American accent voices work in 1.30-1:

 $ espeak -v us-mbrola-3 test
 t 100
 E 157  0 103 100 77
 s 89
 t 77
 _ 263
 _ 1

That looks correct, and the same as eSpeak 1.29.
eSpeak produces phoneme information for mbrola.

To speak with a mbrola diphone voice, pipe the output from espeak into
mbrola. eg:

espeak -v us-mbrola-3 test | mbrola -e /usr/share/mbrola/us3 - out.wav

see http://espeak.sourceforge.net/mbrola.html




-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#439412: missing dependency on jackd

2007-08-27 Thread Jonathan Duddington
On 24 Aug, Thomas K. [EMAIL PROTECTED] wrote:


 After installing with apt-get install espeak running it from the
 command line 

 gives:
 :~$ espeak
 sh: jackd: command not found

 After apt-getting jackd it runs fine, so please add jackd as a
 dependency.

eSpeak does not use jackd.  eSpeak (and myself) know nothing about jack.
I use eSpeak OK here with either portaudio 18 or portaudio 19 without
jack installed.

Perhaps your  portaudio  library has been configured to use jack ?



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#439770: espeak: Up stream is now at version 1.28

2007-08-27 Thread Jonathan Duddington
On 27 Aug, Kenny Hitt [EMAIL PROTECTED] wrote:

 Espeak is now up to version 1.28.  I would like to see this version
 in Debian.

I have now released eSpeak version 1.29, at
http://espeak.sourceforge.net

Fixes include a couple of bugs in earlier versions that could cause
crashes, so eSpeak 1.29 is preferred.



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#415892: espeak: produces incorrect length in wav headers

2007-06-08 Thread Jonathan Duddington
This is really a limitation, rather than a bug.
 
The problem occurs with the  --stdout  option.
Writing to a file using the  -w filename option is OK.

eSpeak doesn't know the length of the speech data when it writes the
header at the start of the output stream.  Unlike writing to a file, it
can't rewind and fill in the length when it's finished.  Therefore the
length fields in the wav header remain at zero.

Should eSpeak omit the wav header completely when writing to stdout? 
Or should it keep the header which contains the sample rate and format
information, so that the output can be automatically fixed by sox?

I'm the eSpeak developer, but I only found this bug report by accident.
 Is there a procedure which passes Debian eSpeak bugs to me, or to the
project bug tracker on SourceForge, so that I can see them?



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#408741: espeak 1.16 doesn't work on powerpc

2007-01-27 Thread Jonathan Duddington
Package: espeak
Version: 1.16-2

A Debian package of the espeak speech synthesizer is available for
powerpc.  I have received reports from users that espeak (versions
earlier than 1.19) does not work on powerpc because of the big-endian
byte ordering.

The same problem will occur or any other big-endian processors.

I am the eSpeak author.  I have now fixed this, and espeak 1.19 is now
reported to run correctly on powerpc.  It is available at
http://espeak.sourceforge.net/ and also as a package for Ubuntu Feisty
(espeak-1.19-0ubuntu1 and espeak-data-1.19-0ubuntu1).

There are two issues:

1.  On powerpc (and presumably other big-endian processors), the byte
order of 16 bit sound samples which are sent to the portaudio sound
output interface is the reverse of that which is written to WAV files.

2.  eSpeak's data files (in directory espeak-data) are compiled data,
and differ from the little-endian versions.  These data files are
compiled from the source data using the espeakedit program, and for
powerpc they must be compiled by running espeakedit on a big-endian
processor.



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]