Festival tips: more understandable and books to audiobooks

Vanquirius · Posted: Tue Sep 21, 2004 9:07 pm Post subject:

Oh, I see. My bad, I thought that was due to lower optimizations, but I see you are right...

MaineZJ · n00b Joined: 03 Dec 2003 Posts: 18

I have recently been trying to use Festival with a project....

I noticed that Festival installs to /usr/share/festival

and when I install mbrola, it's extra voices are installed to /usr/lib/festival

did this to make us1 voice work:
ln -s /usr/lib/festival/voices/english/us1_mbrola/us1 /usr/share/festival/voices/english/us1_mbrola/us1

telegenetic · n00b Joined: 30 Mar 2005 Posts: 2

i was getting this error also i then looked into htsvoice.pl and noticed this line

system("sox -r $SAMPRATE -t raw -s -c 1 -w $WORKDIR/tmp.raw $OUTWAVE");

noticed WORKDIR is tmp
did a USE="-xorg" emerge sox and it works

bitwise · Posted: Mon Jun 13, 2005 2:23 am Post subject:

wow, these sound great, glad I found this thread

Only issue here is the weird seg faults that occur. I still hear the text and everything but I get this:

Wi1d · Posted: Thu Jun 16, 2005 6:56 am Post subject:

mirko_3 · l33t Joined: 02 Nov 2003 Posts: 605 Location: Birreria

I just tried it, I also get this error. Any solution?
_________________
Non fa male! Non fa male!

VinzC · Posted: Sun Jun 26, 2005 9:59 pm Post subject:

Hi all.

I've been able to download french database files (e.g. fr1 - fr7) but I'm lacking the *.scm files (e.g. fr1_mbrola.scm). Does anyone know how/where I can get these kinds of files?

Thanks in advance.
_________________
Gentoo addict: tomorrow I quit, I promise!... Just one more emerge...
1739!

Headhunter123 · Guru Joined: 19 Oct 2002 Posts: 509

This new speech engine works great! Festival is finally understandable, very nice work!

One thing which came up my mind and might be really useful might be a plugin
for firefox, where you select some text and let it read to you by festival.
I usually don't read books at the pc, but i browse a lot of news pages, having them
read to me would be great - anything done like that yet?

telegenetic · n00b Joined: 30 Mar 2005 Posts: 2

I had this working awhile ago, I reinstalled today and it kept seg faulting like others have.
I diff'ed the old text2wave to the new one and have this:

143c143
< (let ((wholeutt (Utterance Text "")))
---
> (let ((wholeutt (utt.synth (Utterance Text ""))))

So on the _new_ text2wave on line 143 i changed:
(let ((wholeutt (utt.synth (Utterance Text ""))))
to
(let ((wholeutt (Utterance Text "")))

No more seg faults.

The whole thing:
tar -xvzf speech_tools-1.2.95-beta.tar.gz
cd speech_tools
./configure
make

cd ..
tar -xvzf festival-1.95-beta.tar.gz
tar -xvzf festlex_CMU.tar.gz
tar -xvzf festlex_OALD.tar.gz
tar -xvzf festlex_POSLEX.tar.gz
tar -xvzf festvox_cmu_us_awb_arctic_hts.tar.gz
tar -xvzf festvox_cmu_us_bdl_arctic_hts.tar.gz
tar -xvzf festvox_cmu_us_jmk_arctic_hts.tar.gz
tar -xvzf festvox_cmu_us_slt_arctic_hts.tar.gz
tar -xvzf festvox_cstr_us_awb_arctic_multisyn-1.0.tar.gz
tar -xvzf festvox_cstr_us_jmk_arctic_multisyn-1.0.tar.gz
tar -xvzf festvox_don.tar.gz
tar -xvzf festvox_ellpc11k.tar.gz
tar -xvzf festvox_en1.tar.gz
tar -xvzf festvox_kallpc16k.tar.gz
tar -xvzf festvox_kallpc8k.tar.gz
tar -xvzf festvox_kedlpc16k.tar.gz
tar -xvzf festvox_kedlpc8k.tar.gz
tar -xvzf festvox_rablpc16k.tar.gz
tar -xvzf festvox_rablpc8k.tar.gz
tar -xvzf festvox_us1.tar.gz
tar -xvzf festvox_us2.tar.gz
tar -xvzf festvox_us3.tar.gz
cd festival
./configure
make
make tests

fix line 143 in bin/text2wave and you should be good.

luca

Massimo B. · Posted: Sat Nov 25, 2006 12:37 am Post subject:

Being on ppc I had to rebuild the binary, then it's executable (at least without argument).
But starting a festival sentence crashes:

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

Hi all,

The HTS working group has released the new version of HMM-based speech synthesis system (HTS version 2.0).
6 new voices (4 male and 2 female) for Festival 1.95 and later were also released.
Some of the speakers in the new voices were the same as those in the previous version, but they are trained with the new technologies.
I think new ones sounds better than the previous version, please try it

You can download them from http://hts.ics.nitech.ac.jp/

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

HTS website moved to http://hts.sp.nitech.ac.jp/

Massimo B. · Posted: Tue Mar 13, 2007 11:11 pm Post subject:

I set up some scripts to issue voice messages.
Now what is the difference between using the festival init script starting a server daemon and using the /dev/speech device OR the usual call approach?

Code:

echo "Lost power. Powersave mode." |festival --tts

_________________
HP ZBook Power 15.6" G8 i7-11800H|HP EliteDesk 800G1 i7-4790|HP Compaq Pro 6300 i7-3770

zaai · Apprentice Joined: 24 Jul 2004 Posts: 175

The CMU ARCTIC Databases, especially the US English slt (female) sounds pretty good. Much better than the pure HTS version of these voices.
Only when words aren't in the dictionary it falls back to diphones which unfortunately adds the nasty hickups.
_________________
* most bugs can be reduced to either dependency or state *

Athlon64 X2 4800+ on Asus M2N SLI deluxe with 4GB Dual channel
video nVidia GForce 7300 GS 512MB (snail); xorg-7.2
kernel 2.6.24-gentoo-r3

vap0rtranz · Posted: Sat Jun 30, 2007 1:26 am Post subject: Voice Skipping Text Files

I'll listen to Salt (voice_cmu_us_slt_arctic_hts) anyday when given the choice between her and Mbrola -- ick.

One issue here. When I pipe text files to festival or when I read files from festival's console, the voice starts fine, then there's two of them talking, and next thing she's skipped a few words. Eitherway:

Massimo B. · Posted: Sat Jun 30, 2007 8:50 am Post subject:

This is also my favorite voice talking to my all the time about new emails, successfull or failed emerges, cron-jobs...
I set it to use alsa:

vap0rtranz · Posted: Sun Jul 01, 2007 7:11 am Post subject:

Massimo B. · Posted: Sun Jul 01, 2007 9:30 am Post subject:

What do you mean about segfaults? Sometimes I get underrun!!! (at least 382,568 ms long), but I think that's my slow machine.
There's only Alsa, no esd running.
_________________
HP ZBook Power 15.6" G8 i7-11800H|HP EliteDesk 800G1 i7-4790|HP Compaq Pro 6300 i7-3770

vap0rtranz · Posted: Mon Jul 02, 2007 4:31 am Post subject:

I mean, if:

vap0rtranz · Posted: Mon Jul 02, 2007 4:46 am Post subject:

starseeker · n00b Joined: 19 Jan 2003 Posts: 40

festival-1.95_beta-r4 USE="X -asterisk -esd -mbrola"

~/.festivalrc:
(voice_rab_diphone)
(set! voice_default 'voice_cmu_us_slt_arctic_hts)

festival --tts test.txt
Duration tree extreme for r 3.64693
Duration tree extreme for r 3.64693
Duration tree extreme for r 3.64693
(successful, plays on speakers.)

./text2wave -o test.wav test.txt
Duration tree extreme for r 3.64693
Duration tree extreme for r 3.64693
Duration tree extreme for r 3.64693
Segmentation fault
(no wav file :-(

failure)

starseeker · n00b Joined: 19 Jan 2003 Posts: 40

An interesting data point - I don't know quite why, or who's fault it is, but when I try to work around this by using vsound the temporary file generated by vsound itself keeps getting re-set in file size:

417816 Jul 21 12:22 vsound20441.au
417816 Jul 21 12:22 vsound20441.au
24600 Jul 21 12:22 vsound20441.au
32792 Jul 21 12:22 vsound20441.au
40984 Jul 21 12:22 vsound20441.au
49176 Jul 21 12:22 vsound20441.au
57368 Jul 21 12:22 vsound20441.au
65560 Jul 21 12:22 vsound20441.au
73752 Jul 21 12:22 vsound20441.au
81944 Jul 21 12:22 vsound20441.au
90136 Jul 21 12:22 vsound20441.au
98328 Jul 21 12:22 vsound20441.au
106520 Jul 21 12:22 vsound20441.au
122904 Jul 21 12:22 vsound20441.au
131096 Jul 21 12:22 vsound20441.au
139288 Jul 21 12:22 vsound20441.au
147480 Jul 21 12:22 vsound20441.au
155672 Jul 21 12:22 vsound20441.au
163864 Jul 21 12:22 vsound20441.au
163864 Jul 21 12:22 vsound20441.au
16408 Jul 21 12:22 vsound20441.au
24600 Jul 21 12:22 vsound20441.au
32792 Jul 21 12:22 vsound20441.au
131096 Jul 21 12:22 vsound20441.au
147480 Jul 21 12:22 vsound20441.au

In the end, all I got in vsound.wav was the last word. Does something about the way this voice works make festival re-set something internally every so often?

vap0rtranz · Posted: Sat Jul 21, 2007 6:13 pm Post subject:

Where did you find those .au files and .wav??
_________________
"Here on the moon, our weekends are so advanced, they encompass the entire week." ~ Ignignokt

starseeker · n00b Joined: 19 Jan 2003 Posts: 40

When I tried the following:

vsound -d festival --tts test.txt

I then opened another xterm and went to that same directory, and watched the temporary au file with ls -l *.au

I used a long enough test file to give me some time to watch.