Howto: Setup more realistic voices in Festival.

[patrock]

The CMU Arctic voices sound magnitudes better than the standard Festival voices. They use prerecorded utterances from Project Gutenberg.

The voices can be downloaded from here. Each voice is around 120 MB. For a description of each voice visit the CMU Arctic page. To hear them visit the voice demo page.

Here's an example of howto install one.

If it's not already installed you'll need to get festlex-cmu:

sudo apt-get install festlex-cmu

Then:

cd /usr/share/festival/voices/english/
sudo wget -c http://www.speech.cs.cmu.edu/cmu_arctic/packed/cmu_us_clb_arctic-0.95-release.tar.bz2
sudo tar jxf cmu_us_clb_arctic-0.95-release.tar.bz2 
sudo ln -s cmu_us_clb_arctic cmu_us_clb_arctic_clunits
sudo cp /etc/festival.scm /etc/festival.scm.backup
sudo echo "(set! voice_default 'voice_cmu_us_clb_arctic_clunits)" >> /etc/festival.scm

Test it out:

echo "This is a test." | festival --tts

You can delete the voice's .tar.bz2 file once it's unpacked.

rm /usr/share/festival/voices/english/cmu_us_clb_arctic-0.95-release.tar.bz2

If you install another one comment out the first voice_default line in /etc/festival.scm with 2 semi-colons.

;;(set! voice_default 'voice_cmu_us_clb_arctic_clunits)

[PinkFloyd102489]

SIOD ERROR: could not open file /usr/share/festival/dicts/cmu/cmulex.scm
closing a file left open: /usr/share/festival/voices/english/cmu_us_clb_arctic_clunits/festvox/cmu_us_clb_arctic_lexicon.scm
closing a file left open: /usr/share/festival/voices/english/cmu_us_clb_arctic_clunits/festvox/cmu_us_clb_arctic_clunits.scm
closing a file left open: /usr/share/festival/init.scm
festival: fatal error exiting.

[PinkFloyd102489]

You also included a female voice in the install process. Might want to post both male and female, as Im now removing the female and installing male.

[patrock]

The male voices sound good too. The HTS voices sound better but I haven't gotten them working yet. You can hear them on the demo page.

[dchurch24]

I cannot get Festival to work at all - it worked fine in Feisty, but in Gusty I get the following error:

SIOD ERROR: could not open file /usr/share/festival/voices/english/cmu_us_clb_arctic_clunits/festvox/cmu_us_clb_arctic_clunits.scm
closing a file left open: /usr/share/festival/init.scm
festival: fatal error exiting.


I have no idea where to go or what to do next! Please help!

[peabody]

First I would try and see if I could get it working with the packaged voices in fiesty. Install the festvox-rablpc16k and put this in your festival.scm:

(set! default_voice 'voice_rab_diphone)

[joerite]

Thanks for this. Is there one similar to GLaDOS. I would make it if I knew how

[unimatrix]

If somebody is getting this error:

lexicon english_poslex not defined

It means you need to:

sudo apt-get install festlex-poslex

Just had to write this down here, because I couldn't find any solutions elsewhere.

[MonkeeSage]

The male voices sound good too. The HTS voices sound better but I haven't gotten them working yet. You can hear them on the demo page.

I got the HTS voices working by doing the following (not sure, but they may depend on the CMU Arctic voices being installed (edit: they don't), which I already have — they are way smaller):

(NOTE: I've started a HOWTO with detailed instructions on how to install all of the available English voices for festival: MBROLA, CMU Arctic, and Nitech HTS. HOWTO.)

Firstly, downloaded them to a tmp dir:

mkdir hts_tmp
cd hts_tmp/
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_awb_arctic_hts-2.0.1.tar.bz2
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_bdl_arctic_hts-2.0.1.tar.bz2
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_clb_arctic_hts-2.0.1.tar.bz2
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_rms_arctic_hts-2.0.1.tar.bz2
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_slt_arctic_hts-2.0.1.tar.bz2
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_jmk_arctic_hts-2.0.1.tar.bz2

Extracted them:

for t in `ls festvox_nitech_us_*` ; do tar xvf $t ; done

Made a dir for them to live in:

sudo mkdir /usr/share/festival/voices/us

Moved them there:

sudo mv lib/voices/us/nitech_us_* /usr/share/festival/voices/us/

Move the hts.scm file into festival's home dir (may not be necessary, but can't hurt):

sudo mv lib/hts.scm /usr/share/festival/hts.scm

Cleaned up:

cd ../
rm -rf hts_tmp/

That's it. :)


Ps. Like I said, I haven't tried them without having the CMU Arctic voices installed as described above, so that may also be a prerequisite. (edit: They work fine by themselves.)

Pss. Festvox also has some HTS voices on their site. I'm not sure, but they look like they may just be copies or older versions of the nitech ones. Not sure though.

[wersdaluv]

I got the HTS voices working by doing the following (not sure, but they may depend on the CMU Arctic voices being installed (edit: they don't), which I already have — they are way smaller):

(NOTE: I've started a HOWTO with detailed instructions on how to install all of the available English voices for festival: MBROLA, CMU Arctic, and Nitech HTS. HOWTO.)

Firstly, downloaded them to a tmp dir:

mkdir hts_tmp
cd hts_tmp/
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_awb_arctic_hts-2.0.1.tar.bz2
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_bdl_arctic_hts-2.0.1.tar.bz2
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_clb_arctic_hts-2.0.1.tar.bz2
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_rms_arctic_hts-2.0.1.tar.bz2
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_slt_arctic_hts-2.0.1.tar.bz2
wget http://hts.sp.nitech.ac.jp/archives/2.0.1/festvox_nitech_us_jmk_arctic_hts-2.0.1.tar.bz2

Extracted them:

for t in `ls festvox_nitech_us_*` ; do tar xvf $t ; done

Made a dir for them to live in:

sudo mkdir /usr/share/festival/voices/us

Moved them there:

sudo mv lib/voices/us/nitech_us_* /usr/share/festival/voices/us/

Move the hts.scm file into festival's home dir (may not be necessary, but can't hurt):

sudo mv lib/hts.scm /usr/share/festival/hts.scm

Cleaned up:

cd ../
rm -rf hts_tmp/

That's it.


Ps. Like I said, I haven't tried them without having the CMU Arctic voices installed as described above, so that may also be a prerequisite. (edit: They work fine by themselves.)

Pss. Festvox also has some HTS voices on their site. I'm not sure, but they look like they may just be copies or older versions of the nitech ones. Not sure though.

Man, this is great! Can you suggest more voices?