Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
Festival tips: more understandable and books to audiobooks
View unanswered posts
View posts from last 24 hours

Goto page 1, 2, 3, 4  Next  
Reply to topic    Gentoo Forums Forum Index Documentation, Tips & Tricks
View previous topic :: View next topic  
Author Message
fireboy1919
n00b
n00b


Joined: 28 May 2002
Posts: 47

PostPosted: Thu Jul 08, 2004 9:45 pm    Post subject: Festival tips: more understandable and books to audiobooks Reply with quote

If you're reading this, you probably know what Festival is - it's a text to speech engine that produces choppy speech from text.

Well, it's been given a major improvement in the form of a hidden Markov Model based synthesis engine.
http://hts.ics.nitech.ac.jp/

These things sound WAY better.

To install:
After installing festival, download and unzip one of the festival engines.
http://hts.ics.nitech.ac.jp/download.html

It's going to create a directory structure that's different than the one currently used by festival, so you have to copy it out to the right spot.
At the time of this writing, english Festival voices are stored in:
/usr/lib/festival/voices/english
by default. Untar and copy the code in lib/voices/us there:
I prefer the "slt" voice.
Code:

tar xzvf cmu_us_slt_arctic_hts.tar.gz
cp -r lib/voices/us/* /usr/lib/festival/voices/english


Making this the default voice:
Unfortunately there's something wrong with these engines, and I didn't want to read all the way through all of Festival to figure it out. Suffice to say it's something in the initialization. You'll have to start with some other voice before loading any of these.
For that reason, make this your /usr/lib/festival/siteinit.scm (with suitable modification based on whichever voice you downloaded) file:
Code:
(voice_kal_diphone)
(set! voice_default 'voice_cmu_us_slt_arctic_hts)


Troubleshooting
1) You have to have perl on your system, because the interface between Festival and this voice is written in perl. Make sure you have perl installed.
2) The actual engine, which is stored at
/usr/lib/festival/voices/english/[whatever you installed]/hts_engine
is an ELF x86 binary. If you don't have that, you'll have to recompile and copy it in there. It's available in the download section from the website I mentioned.

Making audio books
The result is, in fact, so legible that I'd use it to listen to books-on-mp3.
You can get plenty of books from the Gutenberg archive worth listening to for those long drives.

The problem, though, is that festival's text2wave engine tries to keep everything in memory, and I just don't have 10GB of memory to spare.

So...here's a bit of perl to do it in chunks (Yeah, I could have used bash, as the hack I wrote is almost a shell script, but I like perl, okay? Live with it.). It uses lame, and joins the files together in the traditional method with mp3s: using cat.


Code:
#!/usr/bin/perl
open(FILE1,$ARGV[0]);
my ($parnumber,$f)=(0,0);
my $paragraph=50;
while($f=<FILE1>)
{
 for(my $i=0; $i<$paragraph; $i++)
 { last if(!($f .=<FILE1>)); };
 print $parnumber . "\n";
 my $filename=sprintf("temp%05d",$parnumber);
 open(TEXTIN,"|text2wave -F 44100 -o $filename.wav");
 print TEXTIN $f;
 close(TEXTIN);
 system("lame $filename.wav $filename.mp3");
 $parnumber++;
 system("touch $ARGV[1] && cat $filename.mp3 >> $ARGV[1] && rm $filename.*");
};


There's one tweakable value: the variable paragraph. It specifies how many lines to send to text2wave at a time. More means you need more temp space, but then again, there's less overhead from stopping and restarting the program.

Just give it the input plain text filename, and the output mp3 name, and wait. If it's a good-sized novel, you'll be waiting quite some time.

Edit:
Thanks for the catch there, ootneg. I missed that last } when I copied the file to here. There's something else you might use.

A lot of books you get have built-in word wraps. This sounds really bad when it happens in the middle of the word. Well, with the help of perl and ispell, you can get rid of those pesky word wraps. Because I use this so that I can put books on my PDA and have the line wrap look right, it will actually wrap anytime there isn't two "\n" in a row following a period (or a period then a quote mark).

Code:
#!/usr/bin/perl
use Lingua::Ispell qw(spellcheck);
#Lingua::Ispell::allow_compounds(1);
my ($line,$line2,$output);
open(FILE,$ARGV[0]);
$endword,$currline;
while(!($endword eq "eof13432"))
{
 ($currline,$endword)=addline($endword);
 print $currline;
}
close(FILE);
sub addline
{
 my $endword=shift;
 my @word;
 chomp($endword);
 my $line;
 return ($endword . "\n","eof13432") if(!($line=<FILE>));
 return ($endword ,"") if($line=~/^\s*$/ && (!($endword=~/(.*)\.\"?$/)));
 return ($endword . "\n\n","") if($line=~/^\s*$/);
 $line=~/^\s*([^\s]+)(.*?)([^\s]+)?\s*$/;
 my ($beginword,$middle,$end)=($1,$2,$3);
 if($endword=~/(.*)\-$/)
 {
  $endword=$1;
  my $maybe=$endword.$beginword;
  @word=@{spellcheck($maybe)};
  return ($maybe . $middle, $end)  if(!spellcheck($maybe));
 }
 return ($endword . " " . $beginword . $middle,$end) if(length($endword)>0);
 return ($beginword . $middle , $end);
}


To use this piece of code, you have to have Lingua::Ispell installed, as well as ispell (it combines words if it notices that theres a "-" at the end of a line and the combination forms a word in the ispell dictionary).

Check if you have ispell with emerge (emerge -s ispell). Check if you have Lingua::Ispell by using CPAN. To do this, type
Code:
perl -MCPAN -e shell

as root, and then (after configuring it if you haven't done it before), type
Code:
install Lingua::Ispell

at the cpan prompt. Typing exit will exit the CPAN shell.


Last edited by fireboy1919 on Wed Jul 21, 2004 3:36 am; edited 2 times in total
Back to top
View user's profile Send private message
ootneg
Tux's lil' helper
Tux's lil' helper


Joined: 12 Jul 2003
Posts: 82

PostPosted: Fri Jul 09, 2004 5:50 am    Post subject: Re: Festival tips: more understandable and books to audiobo Reply with quote

fireboy1919 wrote:
http://hts.ics.nitech.ac.jp/

fireboy1919 wrote:
http://hts.ics.nitech.ac.jp/download.html


Unfortunately, both of these links timeout when I try to load them
Epiphany-1.2.2 wrote:
Proxy Error
The proxy server received an invalid response from an upstream server.

The proxy server could not handle the request GET http://hts.ics.nitech.ac.jp/.

Reason: Could not connect to remote machine: Operation timed out


Is the site unreliable, or did it disappear just hours after the original post?
Back to top
View user's profile Send private message
floam
Veteran
Veteran


Joined: 27 Oct 2002
Posts: 1067
Location: Vancouver, WA USA

PostPosted: Fri Jul 09, 2004 6:00 am    Post subject: Reply with quote

Me too. Times out.
_________________
Think about your breathing.
http://floam.sh.nu
Back to top
View user's profile Send private message
fireboy1919
n00b
n00b


Joined: 28 May 2002
Posts: 47

PostPosted: Fri Jul 09, 2004 6:55 pm    Post subject: Odd behaviour Reply with quote

I don't know what that's all about.

I don't own the site or control it. You might notice it's in Japan?
Perhaps there's some kind of blockage there.

I just did this, right before posting. I have no idea why it would be down so suddenly.

Edit
As you wish, I've uploaded the best of the voices to my site, as their license for use is incredibly liberal.
I'm low on bandwidth, and running dialup, so you'll have to wait for Japan's servers to come back up to get the rest.
http://rustyp.freeshell.org/gentoo/cmu_us_slt_arctic_hts.tar.bz2

By the way, have you tried messing with Festival? There are a lot better voices than the default one. I like the English guy best of the defaults, whichever one that was. You can see available voices (and the path) by typing
(list voice-locations)

Edit2
For a limited time, also available from me is HTS-1.1.1 for HTK, which is the engine that actually makes the words. You won't need this unless you can't use x86 ELF binaries.
http://rustyp.freeshell.org/gentoo/HTS-1.1.1_for_HTK-3.2.1.tar.gz


Last edited by fireboy1919 on Sat Jul 10, 2004 10:11 am; edited 2 times in total
Back to top
View user's profile Send private message
ootneg
Tux's lil' helper
Tux's lil' helper


Joined: 12 Jul 2003
Posts: 82

PostPosted: Fri Jul 09, 2004 10:18 pm    Post subject: Reply with quote

Yes, I did notice that it is in Japan. I did not expect that it was your site. I also noticed that it had only been a few hours between your post and my attempt. I doubt that they blocked it as a result of high traffic because of this forrum page as only 12 visitors had viewed this post before I did.

Do you still happen to have the tar.gz file? Do you have any way to post it (if their licensing will allow)? I'ld like to try this out. The default voice is not even as good as the one on "War Games".

Thank you.
Back to top
View user's profile Send private message
ootneg
Tux's lil' helper
Tux's lil' helper


Joined: 12 Jul 2003
Posts: 82

PostPosted: Sat Jul 10, 2004 8:38 am    Post subject: Reply with quote

I had not tried to use the other voices. I understood that there was only one voice and that your instructions were to add new voices. I was wrong. I've been reading the documentation and have learned to do quite a bit with this program. I have used all of the voiced that come with it. I also emerged mbrola and used the three voices with that package.

Unfortunately, I have not yet gotten this to work with the voice I got from you. When I try to use it, I get an error:
Quote:
estival> (voice_cmu_us_slt_arctic_hts)
cmu_us_slt_arctic_hts
festival> (SayText "Good morning, welcome to Festival")
Cannot open file /tmp/est_15417_00000/utt.wav as tokenstream
Wave load: can't open file "/tmp/est_15417_00000/utt.wav"
Cannot load wavefile: /tmp/est_15417_00000/utt.wav
#<Utterance 0x4072b818>
festival>


I've got a long drive tomorrow and I should be sleeping already. I'll try getting this to work later this weekend. Hopefully it is just something simple that I have missed.

Thank you for your instructions. I enjoy playing with this kind of stuff.
Back to top
View user's profile Send private message
floam
Veteran
Veteran


Joined: 27 Oct 2002
Posts: 1067
Location: Vancouver, WA USA

PostPosted: Sat Jul 10, 2004 9:04 am    Post subject: Reply with quote

I get the same thing.
_________________
Think about your breathing.
http://floam.sh.nu
Back to top
View user's profile Send private message
banadushi
n00b
n00b


Joined: 26 Jun 2003
Posts: 37
Location: San Antonio, TX USA Earth

PostPosted: Sat Jul 10, 2004 9:55 am    Post subject: Reply with quote

Seems to me, by the name given in the documentation, that these are just the CMU ARCTIC from http://www.festvox.org/cmu_arctic/

[Yea so i'm an idiot, jsut ignore this post]


Last edited by banadushi on Sat Jul 10, 2004 9:57 am; edited 1 time in total
Back to top
View user's profile Send private message
fireboy1919
n00b
n00b


Joined: 28 May 2002
Posts: 47

PostPosted: Sat Jul 10, 2004 9:56 am    Post subject: Possible problems Reply with quote

Perceptive. These people used the arctic dataset to construct their model, but the synthesis engine is different. As I said, they use a Hidden Markov Model.

Like I said, this voice uses a separate synthesis engine, called "HTS."

If you've followed my instructions, it should be
/usr/lib/festival/voices/english/cmu_us_slt_arctic_hts/hts/hts_engine

There's two things I suppose I should mention (will mention above)
1) This is a binary. Specifically, it's an ELF-x86 binary. Source code is available in the download section, so if it doesn't work for you, you can get it there...when the site goes up. I'll put another temp link.
2) The interface between this binary and festival is perl, so you need perl on your system.
Back to top
View user's profile Send private message
ootneg
Tux's lil' helper
Tux's lil' helper


Joined: 12 Jul 2003
Posts: 82

PostPosted: Tue Jul 13, 2004 6:31 am    Post subject: Reply with quote

I did follow the instructions and to have the engine in the same location that you indicated:
Quote:
# ll /usr/lib/festival/voices/english/cmu_us_slt_arctic_hts/hts/hts_engine
-rwxr-xr-x 1 root root 1662248 Jul 10 01:21 /usr/lib/festival/voices/english/cmu_us_slt_arctic_hts/hts/hts_engine
#


I also have perl:
Quote:
% which perl
/usr/bin/perl
% perl -v

This is perl, v5.8.2 built for i686-linux
.
.
.


Apparently the problem is somewhere else. I'll keep looking.
Back to top
View user's profile Send private message
kpoman
Apprentice
Apprentice


Joined: 15 May 2003
Posts: 209
Location: Buenos Aires, Argentina

PostPosted: Tue Jul 13, 2004 6:40 pm    Post subject: Reply with quote

Iam higly interested in setting this up too :)
The site is still dead ! I guess I'll have to wait :oops:
_________________
please, help me, pity on me :'(
Back to top
View user's profile Send private message
ootneg
Tux's lil' helper
Tux's lil' helper


Joined: 12 Jul 2003
Posts: 82

PostPosted: Wed Jul 14, 2004 8:01 pm    Post subject: Reply with quote

I tried this again on a different system and all works well. I followed the exact same steps and everything works. Myabe I had a corrupted download on the other system. I'll have to try to redoo the setup there and see if it makes a difference.

Anyway, the only thing I needed in addition to the emerged packages was the cmu_us_slt_arctic_hts.tar.bz2 package from http://hts.ics.nitech.ac.jp/ offered by fireboy1919. I unpacked the archive and copied it acording to his instructions. On the second computer, all worked well.

Also, the script that he gave works extreemly well after making one syntax correction. It returned the error
Quote:
Missing right curly or square bracket at bookscript.pl line 16, at end of line
syntax error at bookscript.pl line 16, at EOF
Execution of bookscript.pl aborted due to compilation errors.


To correct this, I added a curly bracket '}' just before the semicolon at the end of the last line. This closed the while loop and then the script worked fine.

I used it with the other voices for testing. Now that I have this voice, I will use it for the books that I have in txt format.

Thank you fireboy1919 This is a great post!
Back to top
View user's profile Send private message
kpoman
Apprentice
Apprentice


Joined: 15 May 2003
Posts: 209
Location: Buenos Aires, Argentina

PostPosted: Wed Jul 14, 2004 8:51 pm    Post subject: Reply with quote

could someone maybe post the voice somewhere so wa can download it ? it seems like the link is completely broken !
_________________
please, help me, pity on me :'(
Back to top
View user's profile Send private message
ootneg
Tux's lil' helper
Tux's lil' helper


Joined: 12 Jul 2003
Posts: 82

PostPosted: Wed Jul 14, 2004 11:36 pm    Post subject: Reply with quote

The fourth post in this thread is by fireboy1919 and he has been good enough to post this information as a temp solution while the site is down. Look there for the links.

I only needed the link from the first edit. If you do not have the engine, you may need the link from edit2 in that post. If you would like to hear samples of the different voices without the need to install them first, you may check here
http://festvox.org/voicedemos.html

Hope this helps.
Back to top
View user's profile Send private message
iten
n00b
n00b


Joined: 31 Jan 2004
Posts: 36

PostPosted: Thu Jul 15, 2004 8:49 am    Post subject: Reply with quote

Hmm... I get the same problem as all the others.
A clue might be the htsvoice.pl in the same directory as hts_engine?
raven hts # ./htsvoice.pl
awk: fatal: can't open source file `/label-full.awk' for reading (No such file or directory)
sh: line 1: /hts_engine: No such file or directory

is this the interface you were talking about?
It seems that some paths are wrong...

Actually, on closer look, no... the path is the argument.

Never mind, but does anyone have any clues?
If it helps: I have only emerged festival and speechd, no extra voices. Perhaps you need the original CMU ARCTIC?
Argh, too bad their site is down
Back to top
View user's profile Send private message
ootneg
Tux's lil' helper
Tux's lil' helper


Joined: 12 Jul 2003
Posts: 82

PostPosted: Fri Jul 16, 2004 5:18 am    Post subject: Reply with quote

I couldn't find any differences between the files on the working system and the system where this voice did not work. In my search, I decided to check for any broken library dependencies by running 'revdep-rebuild' (supplied by app-portage/gentoolkit from portage) on the non-working system.. The output included this:
Quote:
Checking dynamic linking consistency...
broken /usr/lib/perl5/site_perl/5.8.0/i586-linux/auto/PDL/Slatec/Slatec.so (requires libg2c.so.0)


It then determined that to fix the dependency, it should run:
Quote:
emerge --oneshot --nodeps =dev-perl/PDL-2.4.0


Since we have already discussed the need for perl, I figured that a broken perl dependency may be responsible for this problem. I let it do this emerge. Unfortunately this did *not* fix the problem. I also tried to remerge festival after this to see if it would make any difference. No luck.

Based on the error message, the problem appears to be either with reading or writting to the /tmp directory.

Quote:
festival> (voice_cmu_us_slt_arctic_hts)
cmu_us_slt_arctic_hts
festival> (SayText "help")
Cannot open file /tmp/est_19344_00002/utt.wav as tokenstream
Wave load: can't open file "/tmp/est_19344_00002/utt.wav"
Cannot load wavefile: /tmp/est_19344_00002/utt.wav
#<Utterance 0x407d5118>
festival>


I checked to see if the expected files were there:
Quote:
/tmp% ll | grep est
/tmp%


No files in /tmp/ contain 'est' in their names. Apparently the temp wave files are not being written as needed. I looked at the permissions of /tmp and cannot see any reason that any application (run by any user) should have problems with writting to it
% ll / | grep tmp
drwxrwxrwt 23 root root 4256 Jul 15 21:10 tmp/
%

With these permissions, any user should be able to create his own files as though they were within his own diskspace. The sticky bit makes it so that users cannot tamper with other user's files, but I see no conflict with the file names.

It seams that I am looking in the right direction, but I don't know what is causing this behavior.

If anybody learns more, I'ld like to understand what is happening.
Back to top
View user's profile Send private message
iten
n00b
n00b


Joined: 31 Jan 2004
Posts: 36

PostPosted: Fri Jul 16, 2004 8:24 am    Post subject: Reply with quote

Mine has no problems writing to the tmp directory, because I was fast enough to catch the directory it creates before festival deletes it.
The contents it seems to create (before deleting it) are:

joel@raven joel $ ls /tmp/est_2*
tmp.f0 tmp.lab tmp.mcep tmp.raw utt.feats

Sorry I couldn't get the permissions.
The file it tries to load is /tmp/(number-- looks like the PID)/utt.wav.

But I also have different est files in my tmp directory, maybe left over from before.
These are not directories, mostly just 44 bytes long (one is 80).

The permissions of the directory it creates:
drwxr-xr-x 2 joel games 176 Jul 16 01:15 est_23219_00001

Hope this helped.
Back to top
View user's profile Send private message
ootneg
Tux's lil' helper
Tux's lil' helper


Joined: 12 Jul 2003
Posts: 82

PostPosted: Sat Jul 17, 2004 12:52 am    Post subject: Reply with quote

You're right. While the program is running, the system generates a directory named /tmp/est_PID_xxxxx/, where PID is the process ID of festival, and xxxxx is a number that incraments each time festival synthesizes. Within this directory are various files. None of them are the wav file from the error. This is the error:
Quote:
festival> (SayText "test")
Cannot open file /tmp/est_19344_00013/utt.wav as tokenstream
Wave load: can't open file "/tmp/est_19344_00013/utt.wav"
Cannot load wavefile: /tmp/est_19344_00013/utt.wav
#<Utterance 0x40abc638>
festival>


I looked for the directory and its contents with the two commands in the same line
Quote:
/tmp% ll | grep est ; ll est*
drwxr-xr-x 2 bob users 176 Jul 16 18:37 est_19344_00013/
total 13
drwxr-xr-x 2 bob users 176 Jul 16 18:37 ./
drwxrwxrwt 24 root root 4320 Jul 16 18:37 ../
-rw-r--r-- 1 bob users 0 Jul 16 18:37 tmp.f0
-rw-r--r-- 1 bob users 957 Jul 16 18:37 tmp.lab
-rw-r--r-- 1 bob users 0 Jul 16 18:37 tmp.mcep
-rw-r--r-- 1 bob users 0 Jul 16 18:37 tmp.raw
-rw-r--r-- 1 bob users 968 Jul 16 18:37 utt.feats
/tmp%


All we know from this is that festival is able to write to /tmp/, but that it has errors other than the ones that I was looking for.
Back to top
View user's profile Send private message
iten
n00b
n00b


Joined: 31 Jan 2004
Posts: 36

PostPosted: Sat Jul 17, 2004 3:44 pm    Post subject: Reply with quote

More clues... when I try to say things with a voice that works, the temp files are still created.
Code:
joel@raven joel $ ls -l /tmp|grep est
-rw-r--r--  1 joel games     529856 Jul 17 08:32 est_13504_00002_aud_00001
-rw-r--r--  1 root root          44 Jul 15 01:37 est_19510_00001
-rw-r--r--  1 root root          44 Jul 15 01:37 est_19510_00003
-rw-r--r--  1 root root          44 Jul 15 01:37 est_19510_00005
-rw-r--r--  1 root root          44 Jul 15 01:37 est_19510_00007
drwxr-xr-x  2 root root          80 Jul 15 01:37 est_19510_00008
-rw-r--r--  1 joel games        867 Jul  3 17:30 test.png

And then again, to get a sense of the numbering system:
Code:

joel@raven joel $ ls -l /tmp/est*
-rw-r--r--  1 joel games 529856 Jul 17 08:33 /tmp/est_13504_00003_aud_00002
-rw-r--r--  1 root root      44 Jul 15 01:37 /tmp/est_19510_00001
-rw-r--r--  1 root root      44 Jul 15 01:37 /tmp/est_19510_00003
-rw-r--r--  1 root root      44 Jul 15 01:37 /tmp/est_19510_00005
-rw-r--r--  1 root root      44 Jul 15 01:37 /tmp/est_19510_00007

This is not a directory that it creates, rather what seems to be a text file, however this could be a different stage in the process that I could not catch.
This is what I managed to snag of the file:
Code:
joel@raven joel $ more /tmp/est_13504_00004_aud_00003
NIST_1A
   1024
channel_count -i 1
sample_count -i 264416
sample_rate -i 16000
sample_coding -s3 pcm
sample_n_bytes -i 2
sample_byte_format -s2 01
end_head

There must be more though, because it seemed like a rather big file...
The real site seems to be permanently gone, so it looks like we're on our own. 8O
ootneg:
Did you try packaging up /usr/lib/festival from the working system and transferring it? I'm not sure what difference it would make, but maybe it's worth a shot.
Back to top
View user's profile Send private message
ootneg
Tux's lil' helper
Tux's lil' helper


Joined: 12 Jul 2003
Posts: 82

PostPosted: Sun Jul 18, 2004 7:04 pm    Post subject: Reply with quote

No, I did not do that. The non working system is behind a non routable firewall, so I have to be on that system to look at it. From there, I can access the working system. I compared the files in two side-by-side terminals. All files are the same size and none appear to be missing.

Just to be sure, I'll do that tomorrow when I'm on the other system again. I'll pack up the files from the working system and create an MD5 hash. After I transfer it, I'll check it against the hash and then put it into place. I'll let you know what happens after I've done that.
Back to top
View user's profile Send private message
Dr_Smack
n00b
n00b


Joined: 15 Jul 2004
Posts: 18

PostPosted: Mon Jul 19, 2004 7:12 pm    Post subject: Reply with quote

I'd just like to toss my hat in the ring as someone for whom festival/text2wave errors. I, too, watched the /tmp/est_* directories and had the same 5 files as iten. The file utt.wav never existed.
Back to top
View user's profile Send private message
ootneg
Tux's lil' helper
Tux's lil' helper


Joined: 12 Jul 2003
Posts: 82

PostPosted: Wed Jul 21, 2004 1:57 am    Post subject: Reply with quote

Looks like the site is back up. let's hope that it stays up for a goo long while.

I compared the files that I unpacked on the two different systems. They are identical. Unfortunately, this means that something else is the problem. I'm thinking that it must be something with the system configuration. but I don't know where to begin looking.

Maybe, with the site being up at http://hts.ics.nitech.ac.jp/ we can find something that will help.
Back to top
View user's profile Send private message
fireboy1919
n00b
n00b


Joined: 28 May 2002
Posts: 47

PostPosted: Wed Jul 21, 2004 4:53 am    Post subject: Troubleshooting Reply with quote

I'm not sure what the problem is for you guys...I can't reproduce it.

I am personally using perl .v5.8.2, speech tools .v1.2.3, and festival-1.4.3-r1 on an x86 machine, so you might look over those dependencies.

Make sure that all the files you put in the lib directory are readable by whatever user you run it as, and that hts_engine and htsvoice.pl are executable by whatever user you run it as...I'm guessing you probably checked that, but just in case...

You might try running each of these and making sure it actually runs and doesn't segfault. That would give you a good indication of what is failing.

It's rather obvious that these programs make their own wave files independent of festival (I tried to use them with the "singing" portion of festival to no avail). It would be nice if they made friendly documentation so we could be as well.
Back to top
View user's profile Send private message
ootneg
Tux's lil' helper
Tux's lil' helper


Joined: 12 Jul 2003
Posts: 82

PostPosted: Wed Jul 21, 2004 5:14 am    Post subject: Reply with quote

My system that can use this voice:
Quote:
* dev-lang/perl :
[ I] 5.8.2-r1 (0)

* app-accessibility/speech-tools :
[ I] 1.2.3 (0)

* app-accessibility/festival :
[ I] 1.4.3-r1 (0)

O/S
# uname -r
2.4.20-gentoo-r8

Arch
AMD Athalon 1.2 on ASUS A7A266 MB


My system that cannot use this voice has:
Quote:
* dev-lang/perl :
[ I] 5.8.2-r1 (0)

* app-accessibility/speech-tools :
[ I] 1.2.3 (0)

* app-accessibility/festival :
[ I] 1.4.3-r1 (0)
O/S
# uname -r
2.6.5-gentoo-r1

ARCH
PIII 733 on Intel MB


I'm sure there are other differences and similarities, but the differences shown above are the kernel (2.4 vs 2.6) and the hardware.

Can we eliminate any of these as being causes of the problem? If anybody has this working on gentoo-dev-sources (2.6 kernel), then we can eliminate the kernel as the cause. If someone has compiled this with
Code:
CFLAGS="-march=pentium3 . . .

and it works, we can eliminate this as being the cause. I'm thinking of recompiling with less optomization in my CFLAGS if nobody can eliminate this as a cause first.

If someone else has other ideas as to what we can compare, please speek up.
Back to top
View user's profile Send private message
fireboy1919
n00b
n00b


Joined: 28 May 2002
Posts: 47

PostPosted: Wed Jul 21, 2004 8:42 am    Post subject: Kernel is not the difference Reply with quote

I can run this program on both my test and server systems.

Server is running 2.4.23, and the test is running 2.6.6. A kernel problem would certainly be very difficult to track down. It's agood thing it's probably not that.

I'm assuming you've tried unmerging and remerging festival?
Back to top
View user's profile Send private message
Display posts from previous:   
Reply to topic    Gentoo Forums Forum Index Documentation, Tips & Tricks All times are GMT
Goto page 1, 2, 3, 4  Next
Page 1 of 4

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum