Festival tips: more understandable and books to audiobooks

Dr_Smack · n00b Joined: 15 Jul 2004 Posts: 18

Well, I'm remerging with very tame CFLAGS now to see if that makes a difference.
I'm running:

2.6.7-gentoo-r10
perl-5.8.4
speech-tools 1.2.3
festival 1.4.3-r1

all on a Celeron 600. I'll edit this post when it finishes with my results (I'm not holding out much hope).

*EDIT* Nope, didn't change a thing.

*EDIT #2* I've now tried installing and running this on 3 different computers and have had the same results (see previous post) on all 3.

chrisashton84 · Posted: Wed Jul 21, 2004 11:11 pm Post subject:

I currently get the same error as ootneg and floam.

ootneg · Tux's lil' helper Joined: 12 Jul 2003 Posts: 82

I recompiled (even unmerged before the new emerge) with no optomizations in my CFLAGS. This made no difference. I replaced both speech-tools and festival this way.

Well . . . Back to the old drawing board.

Chaosite · Posted: Thu Jul 22, 2004 9:04 am Post subject:

Hey, this works great for me!

Thanks!

tridy · n00b Joined: 23 May 2004 Posts: 41 Location: Australia

getting the same error;

festival> (SayText "help")
Cannot open file /tmp/est_13432_00000/utt.wav as tokenstream
Wave load: can't open file "/tmp/est_13432_00000/utt.wav"
Cannot load wavefile: /tmp/est_13432_00000/utt.wav
#<Utterance 0x2a96652850>

there seems to be a problem with speech-tools package, don't know if the problem is during compiling or the actual package but all(or lots of) the scripts/binaries are pointing to PORTAGE_TMPDIR.

for example /usr/lib/speech-tools/bin/tilt_synthesis
includes
extend LD_LIBRARY_PATH "/home/tmp/portage/portage/speech-tools-1.2.3/work/speech_tools/lib:"

I don't know where this should be pointing, there is about 90 similar examples in /usr/lib/speech-tools. I'm hoping this might help someone figure this out. I'm not certain where to go from here.

Rafje · Posted: Thu Jul 22, 2004 7:22 pm Post subject:

Rafje · Posted: Thu Jul 22, 2004 7:34 pm Post subject:

Bingo!

I noticed that I could select voice_us1_mbrola, voice_us2_mbrola and voice_us3_mbrola even before I had emerged mbrola. But when I did select one of those mbrola voices and let festival say something, I got the same "Cannot load wavefile" error.

So, I emerged mbrola, after which the voice_us?_mbrola voices worked. And lo and behold, the voice_cmu_us_slt_arctic_hts magically worked as well. Yay!

So my suggestion is to emerge mbrola and see if it does the trick for you too.

Cheers

Dr_Smack · n00b Joined: 15 Jul 2004 Posts: 18

ootneg · Tux's lil' helper Joined: 12 Jul 2003 Posts: 82

tridy · n00b Joined: 23 May 2004 Posts: 41 Location: Australia

I emerged mbrola

but with no change.....mmmm.

will have a look at the other error, if Rafje has things pointing to the correct place in speech-tools and mbrola, I'm thinking if I can get things pointing in the right direction it might work.

Dr_Smack and ootneg does your /usr/lib/speech-tools/bin/tilt_analysis and other files in /usr/lib/speech-tools/bin point to the wrong place.
i.e.
extend LD_LIBRARY_PATH "/home/tmp/portage/portage/speech-tools-1.2.3/work/speech_tools/lib:"
exec /home/tmp/portage/portage/speech-tools-1.2.3/work/speech_tools/main/tilt_analysis "$@"

I going blindly here, I don't know if this is the problem but it does seem to be an error. Might be a portage problem? i'm running portage-2.0.50-r9.

tridy · n00b Joined: 23 May 2004 Posts: 41 Location: Australia

fixed LD_LIBRARY_PATH problem by updating /usr/portage/app-accessibility/speech-tools. The problem i described seemed to be a known problem.

But I'm still getting

Cannot open file /tmp/est_16977_00000/utt.wav as tokenstream

Rafje was correct.

ootneg · Tux's lil' helper Joined: 12 Jul 2003 Posts: 82

Dr_Smack · n00b Joined: 15 Jul 2004 Posts: 18

chrisashton84 · Posted: Tue Jul 27, 2004 7:10 am Post subject:

Here's my guess, it's just a guess, but I think the wav isn't being created. Since this apparently uses a different engine, has the engine been installed correctly? I can't tell for sure on mine where it should be. Any help here?

Calchan · Posted: Tue Aug 03, 2004 9:49 am Post subject:

I had the same problem. I tried a few things, including recompiling the hts_engine executable, but it didn't work. At least, now, I have an über-optimized hts_engine. Enough with suspense, the solution to our problem is:

tomthewombat · Posted: Tue Aug 17, 2004 12:13 am Post subject:

yeah, the perl script runs sox thought a system() call. would be nice if its left some error messages or a log somewhere when that fails.
_________________
http://www.wombatorium.net

gazoombo · n00b Joined: 22 Feb 2004 Posts: 56 Location: GR. MI. USA

Works for me. ...kinda...
The voices sound fantastic and nothing is broken but I get strange errors mentioning segfaults every time I run festival with the cmu_* voices.

Dr_Smack · n00b Joined: 15 Jul 2004 Posts: 18

tridy · n00b Joined: 23 May 2004 Posts: 41 Location: Australia

emerge sox gets it working for me

thanks

talks a bit fast though, will have to get it to slow down!

crazycat · Posted: Tue Aug 24, 2004 3:51 am Post subject:

Works awesome here on amd64. Thanks for the tipp, i had to replace hts_engine with amd64 compiled one.
saytext works pretty good

- i tried some dumb things with this f*** word and laughed alot. But i got the following error:

kpoman · Posted: Thu Sep 02, 2004 2:33 am Post subject:

hello again!
i am trying to make this thing work but am also getting the segfault:

Heiga Zen · n00b Joined: 13 Sep 2004 Posts: 4 Location: Japan

Hi, my name is Heiga Zen.
I am maintaining and developping HMM-based speech synthesis toolkit.

Today I visited this forum for the first time.
I was very surprised

I'd like to commit to this forum.
If you are interested in developping HMM-based speech synthesis system,
please join our hts-users mailing list.

Next version of festival (2.0?) will include HTS as a module.
Curret HTS voices requires sox, and perl.
I think it causes a lot of problems for you.
I'm very sorry :cry:

But next version does not require them.
Festival 1.95beta has been released already.
(http://www.festvox.org/festival/index.html)
(HTS voices are included initially)
I recommend you to try to use it.
Please give me your feedback.

Thank you very much

P.S. Our web server crushed from July 6th to 21st.
I'm sorry for your inconvienience :cry:

tridy · n00b Joined: 23 May 2004 Posts: 41 Location: Australia

lisa

Vanquirius · Posted: Sun Sep 19, 2004 8:04 pm Post subject: Re: Thank you very much