Sumalatha Emmela | 25 Jul 08:31 2014
Picon

(unknown)



Hi,
     I am working on HMM based Speech Synthesis. We do not have any particular documentation regarding HMM Based Speech Synthesis System.So, i followed the procedure given below

First i have installed all the necessary tools and generated utterances in the Festival folder.

     Tools installed:
 
                  speech_tools-2.1.1-current.tar.gz
                  festival-2.1.1-current.tar.gz
                  festvox-2.5.3-current.tar.gz
        HTK-3.4.1.tar
        HDecode-3.4.1.tar.gz
        HTS-2.2_for_HTK-3.4.1.tar.bz2
        SPTK-3.5.tar.gz
Next i have used the command "$FESTVOXDIR/src/hts_build/build_hts setup" for setup

 After that  i have followed the script file "build_hts"  which is in the "bin" folder. (hts_build/bin/build_hts)

In that script file, i have executed three steps (given below)
./hts_build/bin/build_hts files
./hts_build/bin/build_hts f0      and
./hts_build/bin/build_hts data

After this when i executed the next command "./hts_build/bin/build_hts train"  , i got the following error

 mv: cannot stat
> `/home/idntyti/Desktop/HTS/ou_
us_suma/hts_build/train/trees/db/v/cmp/trees.1':
> No such file or directory
> mv: cannot stat
> `/home/idntyti/Desktop/HTS/ou_us_suma/hts_build/train/trees/db/v/cmp/trees.2':
> No such file or directory
> mv: cannot stat
> `/home/idntyti/Desktop/HTS/ou_us_suma/hts_build/train/trees/db/v/dur/trees.1':
> No such file or directory
> mv: cannot stat
> `/home/idntyti/Desktop/HTS/ou_us_suma/hts_build/train/hmms/db/v/cmp/pdf.1':
> No such file or directory
> mv: cannot stat
> `/home/idntyti/Desktop/HTS/ou_us_suma/hts_build/train/hmms/db/v/cmp/pdf.2':
> No such file or directory
> mv: cannot stat
> `/home/idntyti/Desktop/HTS/ou_us_suma/hts_build/train/hmms/db/v/dur/pdf.1':
> No such file or directory


Kindly give solution for this





Alexander Kivaisi | 23 Apr 10:45 2014
Picon

Fwd: Building a new Voice-Swahili



---------- Forwarded message ----------
From: Alexander Kivaisi <regak2010 <at> gmail.com>
Date: Wed, Apr 23, 2014 at 11:41 AM
Subject: Building a new Voice-Swahili
To: Alan W Black <awb-ETDLCGt7PQU3uPMLIKxrzw@public.gmane.org>, Alok Parlikar <aup+-Z06pOKh3BN4@public.gmane.org.edu>


Dear Alan and Alok,
I hope you still remember me. My name is Alexander Kivaisi, a postgraduate student at University of Cape Town, South Africa. Last year, we had a conversion about creating a festvox voice for Swahili and converting to flite. The conversion did not end since you mention you were trying to improve on how to create new voices for languages such as Swahili. So I was wondering how far are you in doing that. 

Regards,
Alexander

Mousa Moradi | 22 Oct 13:39 2013
Picon

how to use cst_lts_rewrites

Hello all,

I'm going to repeat this question (http://marc.info/?l=festival-talk&m=106322071529973&w=2) which was left unanswered and I have the exact same question.

After using make_lts_rewrite.scm, two files are created that contain the hand-written LTS rules. But now how are we supposed to use them?  The cst_lexicon has no member to hold them and the nearest thing it has, is the lts_rules_set which is of type cst_lts_rules and cannot hold cst_lts_rewrites!

Regards,
--
Mousa Moradi
srikanth ronanki | 16 Oct 08:21 2013
Picon

Differences in Lexicon between festival and flite

Hi,

How is pronunciation varied in festival and flite for building voices with ldom in festvox. 

For example:

insert -> ih n s er t (pronunciation in festival/lib/dicts/cmu)
insert -> ih n s e t  (where is this coming from ??, if it uses the same)

Sometimes, word "the" is pronounced as "dh ax" and if word "the" is placed before words starting with vowels, then it is pronounced as "dh iy" .

Where are these key differences made in flite ? Since, I built the voice with ldom in festvox which uses festival cmulex and now I am unable to synthesize the same using flite 1.5-beta. 

Flite version being used: Flite 1.5-beta.

Please let me know if I am wrong.

--

Thanks & Regards
Srikanth Ronanki,
Speech & Vision Lab,
IIIT-Hyderabad.
Ph.No: +91-9912029220
http://researchweb.iiit.ac.in/~srikanth.ronanki
srikanth ronanki | 16 Oct 08:05 2013
Picon

dictionary in festival and flite

Hi,

How is pronunciation varied in festival and flite for building voices with ldom in festvox. 

For example:

insert -> ih n s er t (pronunciation in festival/lib/dicts/cmu)
insert -> ih n s e t  (where is this coming from ??, if it uses the same)

Sometimes, word "the" is pronounced as "dh ax" and if word "the" is placed before words starting with vowels, then it is pronounced as "dh iy" .

Where are these key differences made in flite ? Since, I built the voice with ldom in festvox which uses festival cmulex and now I am unable to synthesize the same using flite 1.5-beta. 

Flite version being used: Flite 1.5-beta.

Please let me know if I am wrong.

--

Thanks & Regards
Srikanth Ronanki,
Speech & Vision Lab,
IIIT-Hyderabad.
Ph.No: +91-9912029220
http://researchweb.iiit.ac.in/~srikanth.ronanki
Mousa Moradi | 14 Oct 12:44 2013
Picon

mailing archive not available

Hello,

I just wanted to inform that the archive is not available in the given url (https://lists.berlios.de/pipermail/festlang-talk/) and since I've searched a lot about festival and flite and didn't ran into any mailing archive about it, I assume it's not available at all in anywhere!

The archive surely is a valuable source for people like me who have problems and are looking for information. It also eliminates the need to ask repeating questions and wasting time of subscribers.

So if it could be brought back to public access, I and many other people would be grateful.

Regards,

--
Mousa Moradi
MS Student
Department of Computer Engineering
Sharif University of Technology
homepage: http://ce.sharif.edu/~moradi/
Mousa Moradi | 5 Oct 10:06 2013
Picon

converting festival clunits voice to flite

Hi,

I'm converting a clunits voice to flite. It's for a new language (fa) which is named "ir_fa_odj_clunits".

I see here (http://www.speech.cs.cmu.edu/flite/doc/flite_8.html) that we're supposed to hand convert some scheme files to C code. But I didn't find more precise info on what we're exactly supposed to do, I mean which scheme files should be converted in what way to C codes! So I will appreciate guidance on this matter.

And here's more info on my status:
I've done building the 'lpc', 'mcep', 'sts' and 'idx' successfully. When I try to build the generated flite C code, I see that it tries to link to "-lflite_cmu_fa_lang" and "-lflite_cmu_fa_lex" libraries and include "cmu_fa_lang.h" and "cmu_fa_lex.h" files, but it's obvious that they all does not exist, because they're for my new language.

However, how much I searched, I couldn't find any guidance or examples for creating these libraries.
Is there any guidance on how I must implement them? Or is there any sample of another language that I can see?

The folder structure of the voice is also attached.

Thanks in advance,
--
Mousa Moradi
MS Student
Department of Computer Engineering
Sharif University of Technology
homepage: http://ce.sharif.edu/~moradi/
.
├── bin
│   ├── build_flite
│   ├── contour_powernormalize
│   ├── do_build
│   ├── find_powercontours
│   ├── find_powerfactors
│   ├── get_lars
│   ├── get_wavs
│   ├── make_cmn
│   ├── make_dist
│   ├── make_f0
│   ├── make_labs
│   ├── make_lpc
│   ├── make_mcep
│   ├── make_pm
│   ├── make_pm_fix
│   ├── make_pmlab_pm
│   ├── make_pm_pmlab
│   ├── make_pm_wave
│   ├── prompt_them
│   ├── setup_flite
│   ├── simple_powernormalize
│   ├── sphinx_lab
│   ├── sphinxtrain
│   └── traintest
├── cep
│   ├── S_0001.cep
│   ├── S_0002.cep
│   ├── S_0003.cep
│   ├── S_0004.cep
│   ├── S_0005.cep
│   ├── S_0006.cep
│   ├── S_0007.cep
│   ├── S_0008.cep
│   ├── S_0009.cep
│   ├── ...
│   └── S_0948.cep
├── dic
├── emu
│   ├── lab_hlb
│   └── pm_hlb
├── etc
│   ├── emu_f0.tpl
│   ├── emu_lab.tpl
│   ├── emu_pm.tpl
│   ├── mysilence
│   ├── test1
│   ├── txt.done.data
│   ├── uniphone.data
│   └── voice.defs
├── f0
├── festival
│   ├── accents
│   ├── clunits
│   │   ├── all_Asl.desc
│   │   ├── all.desc
│   │   ├── ir_fa_odj.catalogue
│   │   ├── ir_fa_odj.fileordered.scm
│   │   ├── ir_fa_odj.scm
│   │   └── ir_fa_odj.unitordered.scm
│   ├── coeffs
│   ├── disttabs
│   ├── dur
│   ├── f0
│   ├── feats
│   │   ├── a.feats
│   │   ├── ah.feats
│   │   ├── b.feats
│   │   ├── ch.feats
│   │   ├── d.feats
│   │   ├── ...
│   │   └── zh.feats
│   ├── phrbrk
│   ├── relations
│   │   ├── IntEvent
│   │   ├── Phrase
│   │   ├── Segment
│   │   ├── Syllable
│   │   ├── Target
│   │   └── Word
│   ├── trees
│   │   ├── ah.tree
│   │   ├── a.tree
│   │   ├── b.tree
│   │   ├── ch.tree
│   │   ├── ...
│   │   └── z.tree
│   └── utts
│       ├── S_0001.utt
│       ├── S_0002.utt
│       ├── S_0003.utt
│       ├── S_0004.utt
│       ├── S_0005.utt
│       ├── S_0006.utt
│       ├── S_0007.utt
│       ├── S_0008.utt
│       ├── S_0009.utt
│       ├── ...
│       └── S_0948.utt
├── festvox
│   ├── ACKNOWLEDGEMENTS
│   ├── ANNOUNCE-2.0
│   ├── build_clunits.scm
│   ├── build_st.scm
│   ├── config
│   │   ├── common_make_rules
│   │   ├── config
│   │   ├── config.in
│   │   ├── Makefile
│   │   └── project.mak
│   ├── config.cache
│   ├── config.guess
│   ├── config.log
│   ├── config.status
│   ├── config.sub
│   ├── configure
│   ├── configure.in
│   ├── docbook
│   │   └── ...
│   ├── html
│   │   └── ...
│   ├── install-sh
│   ├── ir_fa_lex.out~
│   ├── ir_fa_odj_clunits.scm
│   ├── ir_fa_odj_duration.scm
│   ├── ir_fa_odj_durdata_Asli.scm
│   ├── ir_fa_odj_durdata.scm
│   ├── ir_fa_odj_f0model.scm
│   ├── ir_fa_odj_intonation.scm
│   ├── ir_fa_odj_lexicon2.scm
│   ├── ir_fa_odj_lexicon.scm
│   ├── ir_fa_odj_other.scm
│   ├── ir_fa_odj_phoneset_old.scm
│   ├── ir_fa_odj_phoneset.scm
│   ├── ir_fa_odj_phrasing.scm
│   ├── ir_fa_odj_tagger.scm
│   ├── ir_fa_odj_tokenizer.scm
│   ├── Makefile
│   ├── missing
│   ├── mkinstalldirs
│   ├── README
│   └── src
│       ├── db_example
│       │   └── ...
│       ├── diphones
│       │   ├── darpaasmrpa.scm
│       │   ├── darpaaswb.scm
│       │   ├── diphlist.scm
│       │   ├── ja_schema.scm
│       │   ├── make_diph_index
│       │   ├── make_diph_index.sh
│       │   ├── make_diphs_utts.scm
│       │   ├── Makefile
│       │   ├── setup_diphone
│       │   ├── uk_schema.scm
│       │   └── us_schema.scm
│       ├── duration
│       │   ├── dur.feats
│       │   ├── logdurn.scm
│       │   ├── make_dur_model
│       │   └── Makefile
│       ├── eval
│       │   ├── abtest
│       │   ├── Makefile
│       │   ├── mostest
│       │   ├── text2utts
│       │   └── text2utts.scm
│       ├── general
│       │   ├── align_phones
│       │   ├── cmn
│       │   ├── cmn_main.cc
│       │   ├── cmn_main.o
│       │   ├── contour_powernormalize
│       │   ├── display_f0
│       │   ├── display_sg
│       │   ├── do_build
│       │   ├── emu_f0.tpl
│       │   ├── emu_lab.tpl
│       │   ├── emu_pm.tpl
│       │   ├── emu_sg.tpl
│       │   ├── emu_single_f0.tpl
│       │   ├── find_powercontours
│       │   ├── find_powerfactors
│       │   ├── find_unknowns
│       │   ├── find_unknowns.sh
│       │   ├── fix_pm
│       │   ├── fix_pm_main.cc
│       │   ├── fix_pm_main.o
│       │   ├── get_lars
│       │   ├── get_wavs
│       │   ├── guess_voice_defs
│       │   ├── make_cmn
│       │   ├── make.depend
│       │   ├── make_dirs
│       │   ├── make_dist
│       │   ├── make_f0
│       │   ├── Makefile
│       │   ├── make_labs
│       │   ├── make_lpc
│       │   ├── make_pm
│       │   ├── make_pm_epochs
│       │   ├── make_pm_fix
│       │   ├── make_pmlab_pm
│       │   ├── make_pm_pmlab
│       │   ├── make_pm_wave
│       │   ├── phonealign
│       │   ├── phonealign_main.cc
│       │   ├── phonealign_main.o
│       │   ├── phonealign_test
│       │   ├── play_them
│       │   ├── powcont_norm
│       │   ├── powcont_norm_main.cc
│       │   ├── powcont_norm_main.o
│       │   ├── prompt_them
│       │   ├── simple_powernormalize
│       │   ├── smooth_f0
│       │   ├── smooth_f0_main.cc
│       │   ├── smooth_f0_main.o
│       │   └── traintest
│       ├── intonation
│       │   ├── f0.feats
│       │   ├── impose.scm
│       │   ├── make_f0_model
│       │   ├── make_f0_pm
│       │   ├── Makefile
│       │   └── tree_f0.scm
│       ├── ldom
│       │   ├── all.desc
│       │   ├── build_ldom
│       │   ├── build_ldom.scm
│       │   ├── data_select
│       │   ├── jtime.scm
│       │   ├── ldom.desc
│       │   ├── LDOM_ldom.scm
│       │   ├── LDOM-mode.scm
│       │   ├── Makefile
│       │   ├── setup_ldom
│       │   ├── tel.scm
│       │   └── time.scm
│       ├── Makefile
│       ├── prosody
│       │   ├── bin
│       │   ├── build_prosody.scm
│       │   ├── festvox
│       │   ├── make_dirs_prosody
│       │   ├── Makefile
│       │   ├── pm_unfilled
│       │   └── setup_prosody
│       ├── st
│       │   ├── build_st.scm
│       │   ├── Makefile
│       │   ├── sphinx_lab
│       │   └── sphinxtrain
│       ├── unitsel
│       │   ├── build3
│       │   ├── build_clunits_its.scm
│       │   ├── build_clunits.scm
│       │   ├── build_st.scm
│       │   ├── CLUNITS_clunits.scm
│       │   ├── ir_fa_odj_clunits.scm
│       │   ├── ir_fa_odj_duration.scm
│       │   ├── ir_fa_odj_durdata.scm
│       │   ├── ir_fa_odj_f0model.scm
│       │   ├── ir_fa_odj_intonation.scm
│       │   ├── ir_fa_odj_lexicon2.scm
│       │   ├── ir_fa_odj_lexicon.scm
│       │   ├── ir_fa_odj_other.scm
│       │   ├── ir_fa_odj_phoneset2.scm
│       │   ├── ir_fa_odj_phoneset_old.scm
│       │   ├── ir_fa_odj_phoneset.scm
│       │   ├── ir_fa_odj_phrasing.scm
│       │   ├── ir_fa_odj_tagger.scm
│       │   ├── ir_fa_odj_tokenizer.scm
│       │   ├── Makefile
│       │   ├── make_mcep
│       │   ├── setup_clunits
│       │   ├── syllabic.scm
│       │   └── unitsel.desc
│       ├── vox_diphone
│       │   ├── festvox
│       │   │   ├── cmusphinx2_phones.scm
│       │   │   ├── INST_LANG_lex.scm
│       │   │   ├── INST_LANG_phones.scm
│       │   │   ├── INST_LANG_token.scm
│       │   │   ├── INST_LANG_VOX_diphone.scm
│       │   │   ├── INST_LANG_VOX_dur.scm
│       │   │   ├── INST_LANG_VOX_int.scm
│       │   │   ├── INST_uk_VOX_diphone.scm
│       │   │   ├── INST_uk_VOX_dur.scm
│       │   │   ├── INST_uk_VOX_int.scm
│       │   │   ├── INST_uss2_VOX_diphone.scm
│       │   │   ├── INST_uss2_VOX_dur.scm
│       │   │   ├── INST_uss2_VOX_int.scm
│       │   │   ├── INST_us_VOX_diphone.scm
│       │   │   ├── INST_us_VOX_dur.scm
│       │   │   ├── INST_us_VOX_int.scm
│       │   │   └── Makefile
│       │   ├── japanese
│       │   │   └── Makefile
│       │   └── Makefile
│       └── vox_files
│           ├── general
│           │   ├── INST_LANG_VOX_diphone.scm-diphone
│           │   ├── INST_LANG_VOX_duration.scm
│           │   ├── INST_LANG_VOX_durdata.scm
│           │   ├── INST_LANG_VOX_f0model.scm
│           │   ├── INST_LANG_VOX_intonation.scm
│           │   ├── INST_LANG_VOX_lexicon.scm
│           │   ├── INST_LANG_VOX_other.scm
│           │   ├── INST_LANG_VOX_phoneset.scm
│           │   ├── INST_LANG_VOX_phrasing.scm
│           │   ├── INST_LANG_VOX_tagger.scm
│           │   ├── INST_LANG_VOX_tokenizer.scm
│           │   └── Makefile
│           ├── Makefile
│           ├── prompts
│           │   ├── 3ph.data
│           │   ├── date.data
│           │   ├── jtime.data
│           │   ├── Makefile
│           │   ├── num.data
│           │   ├── tel.data
│           │   ├── time.data
│           │   ├── timit.data
│           │   ├── uniphone.data
│           │   └── weather.data
│           ├── speaker.licence
│           ├── uk
│           │   ├── INST_uk_VOX_duration.scm
│           │   ├── INST_uk_VOX_durdata.scm
│           │   ├── INST_uk_VOX_f0model.scm
│           │   ├── INST_uk_VOX_intonation.scm
│           │   ├── INST_uk_VOX_lexicon.scm
│           │   ├── INST_uk_VOX_other.scm
│           │   ├── INST_uk_VOX_phoneset.scm
│           │   ├── INST_uk_VOX_phrasing.scm
│           │   ├── INST_uk_VOX_tagger.scm
│           │   ├── INST_uk_VOX_tokenizer.scm
│           │   └── Makefile
│           └── us
│               ├── INST_us_VOX_diphone.scm-diphone
│               ├── INST_us_VOX_duration.scm
│               ├── INST_us_VOX_durdata.scm
│               ├── INST_us_VOX_f0model.scm
│               ├── INST_us_VOX_intonation.scm
│               ├── INST_us_VOX_lexicon.scm
│               ├── INST_us_VOX_other.scm
│               ├── INST_us_VOX_phoneset.scm
│               ├── INST_us_VOX_phrasing.scm
│               ├── INST_us_VOX_tagger.scm
│               ├── INST_us_VOX_tokenizer.scm
│               └── Makefile
├── festvox_
│   ├── build_clunits.scm
│   ├── build_st.scm
│   ├── cmu_us_bdl_arctic_lexicon.scm~
│   ├── cmu_us_bdl_arctic_phoneset.scm~
│   ├── ir_fa_odj_clunits.scm
│   ├── ir_fa_odj_duration.scm
│   ├── ir_fa_odj_durdata.scm
│   ├── ir_fa_odj_f0model.scm
│   ├── ir_fa_odj_intonation.scm
│   ├── ir_fa_odj_lexicon2.scm
│   ├── ir_fa_odj_lexicon.scm
│   ├── ir_fa_odj_other.scm
│   ├── ir_fa_odj_phoneset2.scm
│   ├── ir_fa_odj_phoneset_old.scm
│   ├── ir_fa_odj_phoneset.scm
│   ├── ir_fa_odj_phrasing.scm
│   ├── ir_fa_odj_tagger.scm
│   └── ir_fa_odj_tokenizer.scm
├── flite
│   ├── flite_main.c
│   ├── flite_voice_list.c
│   ├── ir_fa_odj.c
│   ├── ir_fa_odj_cl_durmodel.c
│   ├── ir_fa_odj_cl_durmodel.h
│   ├── ir_fa_odj_cltrees.c
│   ├── ir_fa_odj_cltrees.h
│   ├── ir_fa_odj_clunits.c
│   ├── ir_fa_odj_clunits.h
│   ├── ir_fa_odj_lpc_000.c
│   ├── ir_fa_odj_lpc_001.c
│   ├── ir_fa_odj_lpc_002.c
│   ├── ...
│   ├── ir_fa_odj_lpc_070.c
│   ├── ir_fa_odj_lpc.c
│   ├── ir_fa_odj_lpc.h
│   ├── ir_fa_odj_mcep_000.c
│   ├── ir_fa_odj_mcep_001.c
│   ├── ir_fa_odj_mcep_002.c
│   ├── ...
│   ├── ir_fa_odj_mcep_070.c
│   ├── ir_fa_odj_mcep.c
│   ├── ir_fa_odj_mcep.h
│   ├── ir_fa_odj_res_000.c
│   ├── ir_fa_odj_res_001.c
│   ├── ir_fa_odj_res_002.c
│   ├── ...
│   ├── ir_fa_odj_res_070.c
│   ├── Makefile
│   └── voxdefs.h
├── group
├── lab
│   ├── S_0001.lab
│   ├── S_0002.lab
│   ├── S_0003.lab
│   ├── S_0004.lab
│   ├── S_0005.lab
│   ├── S_0006.lab
│   ├── S_0007.lab
│   ├── S_0008.lab
│   ├── S_0009.lab
│   ├── ...
│   └── S_0948.lab
├── lar
├── lpc
│   ├── lpc.params
│   ├── S_0001.lpc
│   ├── S_0001.res
│   ├── S_0002.lpc
│   ├── S_0002.res
│   ├── S_0003.lpc
│   ├── S_0003.res
│   ├── S_0004.lpc
│   ├── S_0004.res
│   ├── S_0005.lpc
│   ├── S_0005.res
│   ├── ...
│   ├── S_0948.lpc
│   └── S_0948.res
├── mcep
│   ├── mcep.params.scm
│   ├── S_0001.mcep
│   ├── S_0002.mcep
│   ├── S_0003.mcep
│   ├── S_0004.mcep
│   ├── S_0005.mcep
│   ├── ...
│   └── S_0948.mcep
├── omid_build_all.sh
├── pm
│   ├── S_0001.pm
│   ├── S_0002.pm
│   ├── S_0003.pm
│   ├── S_0004.pm
│   ├── ...
│   └── S_0948.pm
├── pm_lab
├── prompt-cep
│   ├── S_0001.cep
│   ├── S_0002.cep
│   ├── S_0003.cep
│   ├── S_0004.cep
│   ├── ...
│   └── S_0948.cep
├── prompt-lab
│   ├── S_0001.lab
│   ├── S_0002.lab
│   ├── S_0003.lab
│   ├── S_0004.lab
│   ├── ...
│   └── S_0948.lab
├── prompt-utt
│   ├── S_0001.utt
│   ├── S_0002.utt
│   ├── S_0003.utt
│   ├── S_0004.utt
│   ├── ...
│   └── S_0948.utt
├── prompt-wav
│   ├── S_0001.wav
│   ├── S_0002.wav
│   ├── S_0003.wav
│   ├── S_0004.wav
│   ├── ...
│   └── S_0948.wav
├── recording
├── scratch
│   ├── lab
│   └── wav
├── sig2.wav
├── st
│   └── ...
├── sts
│   ├── S_0001.sts
│   ├── S_0002.sts
│   ├── S_0003.sts
│   ├── S_0004.sts
│   ├── ...
│   └── S_0948.sts
├── test1
├── tmp.lab
├── tree.txt
├── versions
├── wav
│   ├── S_0001.wav
│   ├── S_0002.wav
│   ├── S_0003.wav
│   ├── S_0004.wav
│   ├── ...
│   └── S_0948.wav
├── wavn
└── wrd

116 directories, 14285 files
Khue Nguyen | 22 Aug 18:09 2013
Picon

Error when building large unit selection voice

Dear all,
I am building new voice (in unit selection with about 10000 setences)
I met error :

Making unit distance table for a_V_V (19020)

terminate called after throwing an instance of 'std::bad_alloc'

  what():  std::bad_alloc

Aborted (core dumped)

when I build unit selection tree with command :

festival --heap 20000000 -b festvox/build_clunits.scm '(build_clunits "etc/txt.done.data")'

I tried with SIODHEAPSIZE, festival --heap (change from the smallest to the biggest), but it didn't work and I think I have to split the clunit.

I read through file wow_vi_liz_clunits.scm and replace this code :

 ((string-equal "+" (item.feat i "ph_vc"))
      (string-append
       name
       "_"
       (item.feat i "R:SylStructure.parent.stress")
       "_"
       (wow_vi_liz::nextvoicing i)
       "_"
       (wow_vi_liz::previousvoicing i)

by this :

 ((string-equal "+" (item.feat i "ph_vc"))
      (string-append
       name
       "_"
       (item.feat i "R:SylStructure.parent.stress")
       "_"
       (wow_vi_liz::nextvoicing i)
       "_"
       (wow_vi_liz::previousvoicing i)
       "_"
       (wow_vi_liz::randsplit i)

I attached the original file and the edited file.
With this edited file, I can build the unit selection tree successfully, but when I synthesized, I met the error :

CLUNITS : no predicted class for SIL_CVox_V_1

I wonder this error made by "randsplit" function ?
How can I solve this ?
If I don't use this function, how can I split the unit to smaller one ( I splited it with previousvoicing function and not enough)


P/s

Previousvoicing function :
(define (wow_vi_liz::previousvoicing i)
  (let ((nname (item.feat i "p.name")))
    (cond
;     ((string-equal nname "pau")
;      "PAU")
     ((string-equal "+" (item.feat i "p.ph_vc"))
      "V")
     ((string-equal (item.feat i "p.ph_cvox") "+")
      "CVox")
     (t
      "UV"))))     


Randsplit function
(define (wow_vi_liz::randsplit i)
(let ((randomnumber(rand)))
(cond
((> randomnumber 0.5)
"1")
(t
"0"))))


Best regards.

Link attached file :
https://dl.dropboxusercontent.com/u/74764470/wow_vi_liz_clunits_origin.scm
https://dl.dropboxusercontent.com/u/74764470/wow_vi_liz_clunits_edited.scm

--
Khue Nguyen
Malle ram reddy | 15 Aug 19:44 2013
Picon

building prosodic models for sphinx phone labels

hi,
i have generated labels for some speech database by using sphinxtrain and sphinx2.now i need to generate prosodic labels to play the voices with festival.
can any one help me how should i generate prosodic models by sphinx phone labels

thanks

Khue Nguyen | 13 Aug 11:14 2013
Picon

EHMM in detail

Dear all,
I want to know more about EHMM label in festvox but I can't find any.
Could you suggest me some documents ?

Best regards.
--
Khue Nguyen
Khue Nguyen | 9 Aug 10:30 2013
Picon

SIOD ERROR : ran out of storage when using Text2wave command

Dear all,

I built a new voice in unit selection synthesis.
When I used this command : "text2wave text.txt -o out.wav", I met error : "SIOD ERROR : ran out of storage"
How can I solve this error ?

Best Regards
--
Khue Nguyen

Gmane