Re: eSpeak - PPC architecture
In article [EMAIL PROTECTED], Luke Yelavich [EMAIL PROTECTED] wrote: 2. What happens if you send the speech output to a WAV file? Again, nothing useful. I think eSpeak is actually creating an invalid wav file. I get the same behavior. What architecture and OS is this on? Is the byte order different from i386? Have you ever had eSpeak working on it, or do you know anyone who has? Jonathan, I have come to the conclusion that the phoneme data as well as dictionaries need to be rebuilt, to suit the architecture the package is being built on. I notice that the espeakedit package has the routines to compile the phoneme data, but unfortunately one has to have a GUI running to use the --compile flag of the binary. Are you sure? What happens if you run espeakedit --compile or espeakedit --help ? Is there an error message? I've run my system (MEPIS 6, which is a KDE/Ubuntu distro) in failsafe mode, which I think means no GUI, and I can do espeakedit --compile OK. Is it possible to separate out the phoneme compile routines to a separate binary, independant of any GUI libraries? If so, would you mind sending them to me, so I can use them to ensure all architectures ship with the correct voice and phoneme data? Are you able to build the espeakedit program and compile the espeak-data with it? Does eSpeak then speak OK? The espeakedit program uses wxWidgets data types. It will take a bit of work to convert the phoneme compile routines so that they don't use them. -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
On Sun, Jan 21, 2007 at 09:14:27PM EST, Jonathan Duddington wrote: In article [EMAIL PROTECTED], Luke Yelavich [EMAIL PROTECTED] wrote: 2. What happens if you send the speech output to a WAV file? Again, nothing useful. I think eSpeak is actually creating an invalid wav file. I get the same behavior. What architecture and OS is this on? I got the same behavior on Ubuntu Edgy powerpc. Is the byte order different from i386? I think it is, but I couldn't be sure of that. Have you ever had eSpeak working on it, or do you know anyone who has? Don't know anybody who has had espeak working on powerpc, and never tried it previously. Jonathan, I have come to the conclusion that the phoneme data as well as dictionaries need to be rebuilt, to suit the architecture the package is being built on. I notice that the espeakedit package has the routines to compile the phoneme data, but unfortunately one has to have a GUI running to use the --compile flag of the binary. Are you sure? What happens if you run espeakedit --compile or espeakedit --help ? Is there an error message? [EMAIL PROTECTED]:~/espeak-data$ ../espeakedit-1.18/src/espeakedit --compile Error: Unable to initialize gtk, is DISPLAY set properly? [EMAIL PROTECTED]:~/espeak-data$ ../espeakedit-1.18/src/espeakedit --help Error: Unable to initialize gtk, is DISPLAY set properly? I've run my system (MEPIS 6, which is a KDE/Ubuntu distro) in failsafe mode, which I think means no GUI, and I can do espeakedit --compile OK. The above two commands were run on an i386 Ubuntu Edgy box, with no GUI running, and only minimum libraries installed to build the espeakedit binary. Is it possible to separate out the phoneme compile routines to a separate binary, independant of any GUI libraries? If so, would you mind sending them to me, so I can use them to ensure all architectures ship with the correct voice and phoneme data? Are you able to build the espeakedit program and compile the espeak-data with it? I can build the espeakedit program, but as stated above, I cannot build the espeak data with it. Does eSpeak then speak OK? If I could build the phoneme data successfully, I'd say that it very likely would. I don't have any GUI installed and running on my powerpc machine either, as it is mainly for server use. The espeakedit program uses wxWidgets data types. It will take a bit of work to convert the phoneme compile routines so that they don't use them. Ok I thought as much. -- Luke Yelavich GPG key: 0xD06320CE (http://www.themuso.com/themuso-gpg-key.txt) Email MSN: [EMAIL PROTECTED] Jabber: [EMAIL PROTECTED] signature.asc Description: Digital signature -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
Luke Yelavich, le Sun 21 Jan 2007 22:58:57 +1100, a écrit : Is the byte order different from i386? I think it is, but I couldn't be sure of that. It is different indeed: ppc is big endian. Samuel -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
In article [EMAIL PROTECTED], Luke Yelavich [EMAIL PROTECTED] wrote: I got the same behavior on Ubuntu Edgy powerpc. What happens if you run espeakedit --compile or espeakedit --help ? Is there an error message? [EMAIL PROTECTED]:~/espeak-data$ ../espeakedit-1.18/src/espeakedit --help Error: Unable to initialize gtk, is DISPLAY set properly? OK. I need to make a command line program with the compile phoneme routines, but which does not use wxWidgets. Do I have a deadline? -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
In article [EMAIL PROTECTED], Al Puzzuoli [EMAIL PROTECTED] wrote: My PPC box is running the GUI; so what would be involved in recompiling the data? Is it just a matter of installing the espeakedit program and issuing a command line? First I must check for other byte-order dependencies, and make some changes to the espeakedit program. I will let you know when it's ready to try. You will need to compile the espeakedit program, which shouldn't be a problem. You'll need the wxWidgets libraries, but these may already be installed if you are running Ubuntu. You need the packages: libwxgtk2.6 and libwxgtk2.6-dev. Once it's compiled, hopefully you will be able to just do: epseakedit --compile and eSpeak will then be able speak OK. -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
In article [EMAIL PROTECTED], Jonathan Duddington [EMAIL PROTECTED] wrote: First I must check for other byte-order dependencies, and make some changes to the espeakedit program. I will let you know when it's ready to try. Once it's compiled, hopefully you will be able to just do: epseakedit --compile and eSpeak will then be able speak OK. I think the espeakedit program may be able to run on PPC and compile the phoneme data. espeakedit --compile will compile both the phoneme data and the dictionary data, if it finds the source data in the expected places, for which the defaults are: espeak-data/phsource espeak-data/dictsource Let me know whether it works. -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
In article [EMAIL PROTECTED], Luke Yelavich [EMAIL PROTECTED] wrote: If Al can get the data compiled and sent to me, that should be good enough for now, as I don't see the chances of espeak being used on ia64 and sparc being very high, but for the future, we need a more perminant solution that will allow all architectures to have the correct byte ordering. I've seen reports of eSpeak working OK on ia64. The data should be binary compatible with i386. For example, here is a reported bug which I fixed for version 1.17 http://sourceforge.net/forum/message.php?msg_id=3991770 -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
Upon attempting to compile the phoneme data, I'm getting the following, if there is anything I can do at my end, please let me know: Script started on Sun 21 Jan 2007 08:25:37 PM EST ]0;[EMAIL PROTECTED]: [EMAIL PROTECTED]:~# espeakedit --compile GTK Accessibility Module initialized Compiling phoneme data: /home/alpuzz/espeak-data/phsource/ Unknown phoneme table: 'en' Unknown phoneme table: 'ru' Unknown phoneme table: 'es' Unknown phoneme table: 'cy' Unknown phoneme table: 'nl' Unknown phoneme table: 'pl' Unknown phoneme table: 'de' Unknown phoneme table: 'vi' Unknown phoneme table: 'hi' Unknown phoneme table: 'af' Unknown phoneme table: 'sv' Unknown phoneme table: 'no' Unknown phoneme table: 'en' Unknown phoneme table: 'ro' Unknown phoneme table: 'eo' Unknown phoneme table: 'fi' Unknown phoneme table: 'pt' Unknown phoneme table: 'it' Unknown phoneme table: 'fr' Unknown phoneme table: 'el' Unknown phoneme table: 'en' Compiled phonemes: 477 errors. See 'error.log'. Dictionary errors: ru 642, es 367, cy 508, nl 393, pl 195, de 2478, vi 476, hi 403, af 2670, sv 680, no 163, en 19361, ro 1074, eo 324, fi 769, pt 1376, it 994, fr 3044, el 541, *** stack smashing detected ***: espeakedit terminated Aborted (core dumped) ]0;[EMAIL PROTECTED]: [EMAIL PROTECTED]:~# exit Script done on Sun 21 Jan 2007 08:26:40 PM EST - Original Message - From: Jonathan Duddington [EMAIL PROTECTED] To: ubuntu-accessibility@lists.ubuntu.com Sent: Sunday, January 21, 2007 7:34 PM Subject: Re: eSpeak - PPC architecture In article [EMAIL PROTECTED], Luke Yelavich [EMAIL PROTECTED] wrote: If Al can get the data compiled and sent to me, that should be good enough for now, as I don't see the chances of espeak being used on ia64 and sparc being very high, but for the future, we need a more perminant solution that will allow all architectures to have the correct byte ordering. I've seen reports of eSpeak working OK on ia64. The data should be binary compatible with i386. For example, here is a reported bug which I fixed for version 1.17 http://sourceforge.net/forum/message.php?msg_id=3991770 -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
On Mon, Jan 22, 2007 at 11:23:39AM EST, Jonathan Duddington wrote: I think the espeakedit program may be able to run on PPC and compile the phoneme data. espeakedit --compile will compile both the phoneme data and the dictionary data, if it finds the source data in the expected places, for which the defaults are: espeak-data/phsource espeak-data/dictsource Let me know whether it works. Ok. The following steps were carried out on two different boxes, one powerpc, and one i386. I managed to get X forwarding set up on my powerpc, so that espeakedit can run. Here is what I did. * Unpacked fresh directories of both zip files, espeak-1.18-source.zip, and espeakedit-1.18.zip. * Created an empty espeak-data directory in my home directory, and copied the voices directory into it, from the espeak-1.18-source/espeak-data directory. * Copied dictsource from espeak-1.18 into espeak-data. * Copied the phsource directory from espeakedit-1.18 into the espeak-data directory I previously created. * Built a fresh binary of espeakedit, and moved it to my home directory. * Entered the espeak-data directory, which as previously stated, only has the voices, phsource, and dictsource directories. * Ran the command: ~/espeakedit --compile Attached are two log files, from both attempted compile runs. Notice the vast difference in output. The other difference is that the *_dict files are built on powerpc, and only the phoneme datafiles are present on i386. However, i386 has correctly built phoneme data, and powerpc doesn't, as evident from the Unknown phoneme table message when compiling, and also when I attempt to make espeak say something. Please let me know if you need any clarrification on what I have outlined above, and I hope these log files are useful. -- Luke Yelavich GPG key: 0xD06320CE (http://www.themuso.com/themuso-gpg-key.txt) Email MSN: [EMAIL PROTECTED] Jabber: [EMAIL PROTECTED] Compiling phoneme data: /home/luke/espeak-data/phsource/ Can't read dictionary file: '/home/luke/espeak-data/en_dict' Can't read dictionary file: '/home/luke/espeak-data/en_dict' Compiled phonemes: 0 errors. Compiled 0 dictionaries Compiling phoneme data: /home/luke/espeak-data/phsource/ Unknown phoneme table: 'en' Can't read dictionary file: '/home/luke/espeak-data/en_dict' Unknown phoneme table: 'af' Can't read dictionary file: '/home/luke/espeak-data/af_dict' Unknown phoneme table: 'cy' Can't read dictionary file: '/home/luke/espeak-data/cy_dict' Unknown phoneme table: 'de' Can't read dictionary file: '/home/luke/espeak-data/de_dict' Unknown phoneme table: 'el' Can't read dictionary file: '/home/luke/espeak-data/el_dict' Unknown phoneme table: 'en' Can't read dictionary file: '/home/luke/espeak-data/en_dict' Unknown phoneme table: 'eo' Can't read dictionary file: '/home/luke/espeak-data/eo_dict' Unknown phoneme table: 'es' Can't read dictionary file: '/home/luke/espeak-data/es_dict' Unknown phoneme table: 'fi' Can't read dictionary file: '/home/luke/espeak-data/fi_dict' Unknown phoneme table: 'fr' Can't read dictionary file: '/home/luke/espeak-data/fr_dict' Unknown phoneme table: 'hi' Can't read dictionary file: '/home/luke/espeak-data/hi_dict' Unknown phoneme table: 'it' Can't read dictionary file: '/home/luke/espeak-data/it_dict' Unknown phoneme table: 'nl' Can't read dictionary file: '/home/luke/espeak-data/nl_dict' Unknown phoneme table: 'no' Can't read dictionary file: '/home/luke/espeak-data/no_dict' Unknown phoneme table: 'pl' Can't read dictionary file: '/home/luke/espeak-data/pl_dict' Unknown phoneme table: 'pt' Can't read dictionary file: '/home/luke/espeak-data/pt_dict' Unknown phoneme table: 'ro' Can't read dictionary file: '/home/luke/espeak-data/ro_dict' Unknown phoneme table: 'ru' Can't read dictionary file: '/home/luke/espeak-data/ru_dict' Unknown phoneme table: 'sv' Can't read dictionary file: '/home/luke/espeak-data/sv_dict' Unknown phoneme table: 'vi' Can't read dictionary file: '/home/luke/espeak-data/vi_dict' Unknown phoneme table: 'en' Compiled phonemes: 477 errors. See 'error.log'. Dictionary errors: af 2670, cy 508, de 2478, el 541, en 19361, eo 324, es 367, fi 769, fr 3044, hi 403, it 994, nl 393, no 163, pl 195, pt 1376, ro 1074, ru 642, sv 680, vi 476, *** stack smashing detected ***: /home/luke/espeakedit terminated Aborted __ Phoneme Table: 'base' 167: File not SPEC or RIFF: 'vowel/@' 175: File not SPEC or RIFF: 'vowel/@-' 188: File not SPEC or RIFF: 'ustop/percus10' 198: File not SPEC or RIFF: 'j/j@' 199: File not SPEC or RIFF: 'j2/j2@' 200: File not SPEC or RIFF: 'w/w@' 201: File not SPEC or RIFF: 'l/l@' 203: File not SPEC or RIFF: 'l^/j2@' 204: File not SPEC or RIFF: 'r/r@' 205: File not SPEC or RIFF: 'r2/r2@' 206: File not SPEC or RIFF: 'm/m@' 207: File not SPEC or RIFF: 'n/n@' 208: File not SPEC or RIFF: 'nn/nn@' 209: File not SPEC or RIFF: 'n^/n^@' 210: File not SPEC or RIFF: 'l/[EMAIL PROTECTED]' 211: File not SPEC or RIFF:
Re: eSpeak - PPC architecture
On Tue, Jan 16, 2007 at 02:30:33PM EST, Al Puzzuoli wrote: Hi Jonathan, I tried the experiments you suggested. 1. What text output do you get if you use the -x option, eg: espeak -x hello? Interestingly, I'm not getting any text at all. When I do this, I'm immediately dropped back to the shell with no output. 2. What happens if you send the speech output to a WAV file? Again, nothing useful. I think eSpeak is actually creating an invalid wav file. I get the same behavior. Jonathan, I have come to the conclusion that the phoneme data as well as dictionaries need to be rebuilt, to suit the architecture the package is being built on. I notice that the espeakedit package has the routines to compile the phoneme data, but unfortunately one has to have a GUI running to use the --compile flag of the binary. Is it possible to separate out the phoneme compile routines to a separate binary, independant of any GUI libraries? If so, would you mind sending them to me, so I can use them to ensure all architectures ship with the correct voice and phoneme data? I don't want to make espeak 1.18 packages available, until we have this sorted out properly. Thanks. -- Luke Yelavich GPG key: 0xD06320CE (http://www.themuso.com/themuso-gpg-key.txt) Email MSN: [EMAIL PROTECTED] Jabber: [EMAIL PROTECTED] signature.asc Description: Digital signature -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
In article [EMAIL PROTECTED], Al Puzzuoli [EMAIL PROTECTED] wrote: After recompiling the dictionary data, I no longer get that error; However, still no speech. If I send a string to eSpeak, it just exits with no noticeable output. Does that suggest incompatible phoneme data? I'm not sure. What text output do you get if you use the -x option, eg: espeak -x hello should give the phoneme codes: [EMAIL PROTECTED]'oU What happens if you send the speech output to a WAV file? Can you then play the WAV file later?, eg: espeak -w test.wav hello Do you know whether anyone has ever had eSpeak working on a similar PPC ? -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
Hi Jonathan, I tried the experiments you suggested. 1. What text output do you get if you use the -x option, eg: espeak -x hello? Interestingly, I'm not getting any text at all. When I do this, I'm immediately dropped back to the shell with no output. 2. What happens if you send the speech output to a WAV file? Again, nothing useful. I think eSpeak is actually creating an invalid wav file. When I attempt to play the file using aplay, I get: Playing WAVE 'test.wav' : Signed 16 bit Little Endian, Rate 576061440 Hz, Mono aplay: set_params:965: Unable to install hw params: ACCESS: RW_INTERLEAVED FORMAT: S16_LE SUBFORMAT: STD SAMPLE_BITS: 16 FRAME_BITS: 16 CHANNELS: 1 RATE: NONE PERIOD_TIME: (21333 21334) PERIOD_SIZE: (12289118 12289119) PERIOD_BYTES: (24578236 24578238) PERIODS: (16 17) BUFFER_TIME: (341333 341334) BUFFER_SIZE: 196628971 BUFFER_BYTES: 393257942 TICK_TIME: 0 I'm not sure where to go with this nextin terms of further troubleshooting. If it would help at all, I could grant you ssh access to the box. Please let me know if that would be of use, or if there is anything else you'd like me to try. --Al -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
In article [EMAIL PROTECTED], Jonathan Duddington [EMAIL PROTECTED] wrote: Delete the file /usr/share/espeak-data/en_dict file, and also make sure you don't have a directory espeak-data in your user's home directory. ... because it will use the espeak-data in the user's home directory in preference to /usr/share Then try re-compiling the dictionary data: espeak --compile=en I should have said, first go into the dictsource directory, which contains the en_rules and en_list files. It compiles en_rules and en_list to make espeak-data/en_dict. Can it speak then? If the phoneme data (espeak-data/phondata,phonindex,phontab) is also incompatible, then that's a bigger problem. If so, you will need espeakedit running on PPC to be able to compile the phoneme data. Which means you need a port of the espeakedit program to PPC. -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility
Re: eSpeak - PPC architecture
Hi, After recompiling the dictionary data, I no longer get that error; However, still no speech. If I send a string to eSpeak, it just exits with no noticeable output. Does that suggest incompatible phoneme data? --Al - Original Message - From: Jonathan Duddington [EMAIL PROTECTED] To: Jonathan Duddington [EMAIL PROTECTED]; ubuntu-accessibility@lists.ubuntu.com Sent: Sunday, January 14, 2007 9:42 PM Subject: Re: eSpeak - PPC architecture In article [EMAIL PROTECTED], Jonathan Duddington [EMAIL PROTECTED] wrote: Delete the file /usr/share/espeak-data/en_dict file, and also make sure you don't have a directory espeak-data in your user's home directory. ... because it will use the espeak-data in the user's home directory in preference to /usr/share Then try re-compiling the dictionary data: espeak --compile=en I should have said, first go into the dictsource directory, which contains the en_rules and en_list files. It compiles en_rules and en_list to make espeak-data/en_dict. Can it speak then? If the phoneme data (espeak-data/phondata,phonindex,phontab) is also incompatible, then that's a bigger problem. If so, you will need espeakedit running on PPC to be able to compile the phoneme data. Which means you need a port of the espeakedit program to PPC. -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility -- Ubuntu-accessibility mailing list Ubuntu-accessibility@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-accessibility