I did everything you've given in this grapheme method. Also I have created the test files. They sound good. I want to provide some new sentences and get the voice for them.
May I know what is the code to provide a new sentence and get its voice output from here?

Thank you.

I tried to follow the steps to create clustergen voice.
However the sequence is failing to generate voice.

The script
./bin/build_cg_rfs_voice
invokes the script
./bin/build_cg_voice

This script invokes the following line
./bin/do_clustergen dur etc/txt.done.data.train

Everything goes fine upto that line however

I start getting errors like
Cannot open file festival/utts_hmm/arctic_a0050.utt as tokenstream
load_utt: can't open utterance input file festival/utts_hmm/arctic_a0050.utt
utt.load: loading from "festival/utts_hmm/arctic_a0050.utt" failed
Cannot open file festival/utts_hmm/arctic_a0048.utt as tokenstream
load_utt: can't open utterance input file festival/utts_hmm/arctic_a0048.utt

etc

I looked into what is being done by above line. It invokes ./bin/do_clustergen. That has following snippet

if [ $1 = "dur" ]
then
./bin/make_dur_model_mcep $PROMPTFILE
cp -p festvox/$FV_VOICENAME"_dur.scm" festvox/$FV_VOICENAME"_durdata_cg.scm"
exit 0
fi
........the above snippet invokes ./bin/make_dur_model_mcep. It has following logic.
...
awk '{printf("festival/utts_hmm/%s.utt\n",$2)}' >utthmmfile
...

however the folder festival/utts_hmm is not populated at all at the time the above code is invoked.
.... and thus the errors start.

Further down the line .........it appears to go through an infinite loop.........not sure about this.

Please let me know if I am following any wrong sequence of steps.
Please direct me to proper source where I can get any guideline, if you are busy.

Thanks in advance.
-Mahesh

if TYPE == tempo use $FESTVOXDIR/src/vc/src/tempo/tempo

$FESTVOXDIR/src/vc/src/tempo/tempo has gone ...

in do_clustergen
FRAMELEN=$(echo | awk "{print int(0.025*$SAMPFREQ)}")
FRAMESHIFT=$(echo | awk "{print int(0.005*$SAMPFREQ)}")

	FWARP[8000]=0.312
	FWARP[11025]=0.357
	FWARP[16000]=0.42
	FWARP[22050]=0.455
	FWARP[32000]=0.504
	FWARP[44100]=0.544
	FWARP[48000]=0.554
    FREQWARP=FREQWARP=${FWARP[$SAMPFREQ]}        

in src/clustergen/get_cd_dtw_mgc we only have the cases of
16000 and 48000

for do_pm -scaleN is fixed to 0.9
but in general/simple_powernormalize
0.65 is chosen ?
Are the different options "historical" 'due to the quality of the rcording at the time ?
What are the impact of the choice made ?

https://github.com/festvox/festvox/blob/3be8dd4e3d2ac680d8d6b33fe6212322636b45c8/src/intonation/make_f0_v_sptk#L126C6-L127C63

doesn't follow the usage of ch_wave:

Usage: ch_wave [input file0] [input file1] ... -o [output file]
Summary: change/copy/combine waveform files
use "-" to make input and output files stdin/out

and lead to the error

Write access failed for temporary file
Cannot open file _tmp.3067154/est_3067160_00000 as tokenstream

Could it be:

`$ESTDIR/bin/ch_wave -otype raw  wav/$fname.wav > $TMP.raw`

(without <)

I can use num2words for automatic numbers words converter. Using (system) function, in my function (define (festpb_pt::number token name)

(define (festpb_pt::number token name)
  "(festpb_pt::number token name)
Return list of words that pronounce this number in pt."

;Set name random, using rand function
(set! ra (rand))

;Set command for system function, piper output of num2words in file with name of variable ra 
(set! command (format nil "num2words -l pt-BR %s > %s" name ra))

;Execute command using system function
(system command)

;Load file how a list with load function
(load ra t)
)

For deletion of cache file have function in Festvox Doc with this in my function (festpb_pt::number token name)

(delete-file ra)

But my function, not return when have this function.

At least for me...FWIW, thanks!

in the doc, there is no output format "5" : only 0 for pitch, 1 for f0, 2 for log(f0)
pitch - pitch extraction
usage:
pitch [ options ] [ infile ] > stdout
options:
-a a : algorithm used for pitch [0]
estimation
0 (RAPT)
1 (SWIPE')
-s s : sampling frequency (kHz) [16.0]
-p p : frame shift [80]
-T T : voiced/unvoiced threshold [0.0]
(used only for RAPT algorithm)
-t t : voiced/unvoiced threshold [0.3]
(used only for SWIPE' algorithm)
-L L : minimum fundamental [60]
frequency to search for (Hz)
-H H : maximum fundamental [240]
frequency to search for (Hz)
-o o : output format [0]
0 (pitch)
1 (f0)
2 (log(f0))

 for i in `seq 1 5`
    do
	$ESTDIR/bin/sigfilter -firfilter $FTMP.h$i -forder $FORDER wav/$fname.wav -otype raw | \
	    $X2X +sf | $SPTKDIR/bin/pitch -a 0 -H $F0MAX -L $F0MIN -p $FRAMESHIFT -s $SAMPKHZ -o 5 | \
	    $X2X +fa4 | awk '{print $4}' > $TMP.h$i.str
    done

luckily ifor swipe (SPTK-3.6/bin/pitch/swipe/swipe.c ) the default is : /* pitch */
but what about rapt ? (pitch -a 0 )

Following the script https://github.com/festvox/festival/files/2300513/fest_build.txt and https://github.com/festvox/festival/issues/13#issuecomment-414084297,I try to build a festvox voice.
Everything goes fine,

until:./bin/do_clustergen parallel cluster etc/txt.done.data.train
The error is:"ERROR: unbound variable : voice_kal_diphone".

I didn't find any related problems on the Internet.Can anyone help solve this problem？

Hello, In building the voice model, I noticed that it has files called.frq what he are? Are files related to frequency and pitch and etc? I have a database that has some files .frq and files .llsm, is it possible to utilize it these files in the database for voice construction?

Please find the logs attached
festvox does not build in my system
Please find ./configure and make logs below.

conf_log.txt
make_log.txt

    F0MIN=50
    F0MAX=200
    F0MEAN=110
    # >....<
    F0_ARGS=`echo $F0MIN $F0MAX $F0MEAN | awk '{printf("-min %f -max %f -def %f",1.0/$2,1.0/$1,1.0/$3)}'`
    PM_ARGS='-wave_end -lx_lf 140 -lx_lo 111 -lx_hf 80 -lx_ho 51 -med_o 0'
    $ESTDIR/bin/ch_wave -scaleN 0.9 wav/$fname.wav -F 16000 | 
    $ESTDIR/bin/pitchmark -o pm_unfilled/$fname.pm -otype est $PM_ARGS $F0_ARGS

will raise

Write access failed for temporary file
Cannot open file TMP/est_868612_00000 as tokenstream
Wave load: can't open file "TMP/est_868612_00000"
Cannot recognize file format or cannot access file: "-"

Could it be something like ?

    # pitchmark has a file name as input
    "$ESTDIR"/bin/ch_wave -scaleN 0.9 -F 16000 wav/"$fname".wav -o tmp_"$fname"
    "$ESTDIR"/bin/pitchmark -o pm_unfilled/"$fname".pm -otype est  $PM_ARGS $F0_ARGS tmp_"$fname"

@saikrishnarallabandi I have been trying to build a synthesizer from these steps

the ./bin/do_build command is not able to find the option get_phseq. I have installed all my libraries using
fest_build.sh.txt

`
(falconenv) gnani@prajwal:~/projects/falcon/festvox/voices/indic/hindi_female/vox$ ./bin/do_build get_phseq

./bin/do_build unknown option get_phseq
`

https://github.com/festvox/festvox/blob/ec3e8b4ee739447b7c2a062506759049d4ab1b5a/src/ehmm/etc/mysp_settings#L1-Ll119

leave alone

Lporder: 12
CepsNum: 16

does it mean that that we don't have to to touch it if we work with other SamplingFreq than 16000? it will help to know what exactly CepsNum is ( it's link with MCEP Order if any ..)

I followed the steps mentioned in Building Indic voices and another issue's comment and tried to build voice for Bengali language, but failed to do so. What I did is given below:

• 1. I downloaded the data available in Fesvox's example Indic database. I used only a small portion of it (20 files) for the process, as training took long time in the first try.
• 2. I installed SPTK-3.6,festvox2.4, festival2.4, speech-tools2.4 in a directory, installed them (without error) and configured FESTVOXDIR, ESTDIR, SPTKDIR environment variables (in Ubuntu 16.04).
• 3. I installed SPTK-3.6,festvox2.5, festival2.5, speech-tools2.5 in a directory, installed them (without error) and configured FESTVOXDIR, ESTDIR, SPTKDIR environment variables (in Ubuntu 16.04).
• 4. I followed the instructions given in the link Building Indic voices, as well as from another issue's comment for both version-2.4 and version-2.5 (I know about the change in these version about setting (defvar lex:language 'Hindi) to Bengali and set them accordingly).
• 5. When building with festvox2.4 and other 2.4-releases, I didn't face problem during training, the whole process went smoothly. But when I ran the following lines mentioned in Building Indic voices, I get error.

And generate some test examples, the first to give MCD and F0D objective measures, the second to generate standard tts output
./bin/do_clustergen cg_test resynth cgp etc/txt.done.data.test
./bin/do_clustergen cg_test tts tts etc/txt.done.data.test
Error I faced:

CG test_resynth bn_00010
SIOD ERROR: unbound variable : rfs_info
CG test_resynth bn_00010 Failed
CG test_resynth bn_00021 
Segmentation fault (core dumped)
awk: cmd. line:2: fatal: division by zero attempted
awk: cmd. line:2: fatal: division by zero attempted
awk: cmd. line:2: fatal: division by zero attempted
awk: cmd. line:2: fatal: division by zero attempted

• 5. When tried to build with festvox-2.5 and other 2.5-releases from this github account, I faced error, after the following command mentioned in both of the sources (another issue's comment and Building Indic voices ). The command is:
     ./bin/do_clustergen parallel cluster etc/txt.done.data.train
     and the error I found:

Setting clustergen params
SIOD ERROR: unbound variable : cg:f0_clustersize
BACKTRACE:
   0: cg:f0_clustersize
   1: (list (quote wagon_cluster_size_f0) cg:f0_clustersize)
   2: (list
    (quote (clunit_relation mcep))
    (list (quote wagon_cluster_size_mcep) cg:mcep_clustersize)
    ...)
   3: (append
    (list
     (quote (clunit_relation mcep))
     (list (quote wagon_cluster_size_mcep) cg:mcep_clustersize)
     ...)
    clunits_params)
   4: (set!
    clustergen_params
    (append
     (...)
     clunits_params))
   5: (begin
    (set! cg:parallel_tree_build t)
    (build_clustergen "etc/txt.done.data.train"))

Please let me know, if I am doing something wrong, and if possible, direct me to the correct steps. If possible, direct me to some proper sources that I can use to build Indic voices in Festival/Festvox.

Thanks,
~Numan

as in
"(INST_LANG_VOX::select_other)
Set up the anything esle for the voice."

+( typo esle -> else )

in src/general/make_f0 we have

$ESTDIR/bin/pda -L -fmin 70 -fmax 240 -o f0/$fname.f0 -otype ssff $PDA_PARAMS wav/$fname.wav

and we have in src/general/prune_middle_silences

PDAPARAMS='-fmin 70 -fmax 240'
followed by
$ESTDIR/bin/pda -L $PDAPARAMS -otype est_ascii wav/$fname.wav | ...

so PDA_PARAMS and PDAPARAMS are not the same .. or ... should be the same

• in make_f0, unless we knew about it the variable $PDA_PARAMS is unset.
• in src/intonation/find_f0_stats or hts_build/build_utts
  are used fmin=$FMIN and fmax=$FMAX as provided by etc/f0.params

Hello,
I built out a new system on Ubunto 20.x. I am trying to do the 'Telling the time' exercise from the documentation. I have run into a problem with the display_sg utility. It returns the following message when I execute the command:

/home/greg/projects/build/festvox/src/general/display_sg: 68: emulabel: not found

I reinstalled the festvox tools and did not see any errors. Does anyone have any suggestions?

I am trying to build a lexicon with pronunciation for Tamil and Telugu using the indic frontend presented in the "The Festvox Indic Frontend for Grapheme-to-Phoneme Conversion" paper.

I found this indic_lexicon.scm file but am not sure how exactly to use it independently just to build a lexicon (word -> phones) without any of the TTS aspects.

Would appreciate inputs on this.

Regards,
Sourya

I'm tring to build Tamil voice . I have followed all the steps mentioned in " https://github.com/festvox/festvox/blob/master/docbook/grapheme.sgml ". But when I'm trying to build prompts error comes as " ta_1871 PROMPTS
LEXICON: Word நீண்ட (plus features) not found in lexicon " for almost all the words.
But all those words are in the lexicon.
Why is that error coming ?

I have already build some indic tts models using festvox, I want to integrate them into festival so that i can play the utterance from command line. Like:
festival '(SayText "नमस्कार आपका स्वागत है.")'.

in /festvox/src/clustergen/get_str_sptk (for example)

MGCEP=$SPTKDIR/bin/mcep
doesn't follow what it seems a pattern
   X2X=$SPTKDIR/bin/x2x
   ..
   SOPR=$SPTKDIR/bin/sopr
but both $SPTKDIR/bin/mcep and $SPTKDIR/bin/mgcep exists

could it be a typo ?

      the option -silences  of  $FESTVOXDIR/src/general/smooth_f0  is a comma separated list of silence names
so I expected  the LIST of phones I declared as silences in my phoneset.scm will be taken into account, after all I'm not te only one to want a list not reduced to a single element  for radio-phones (PhoneSet.silences '(pau h# brth)) 

but indeed we have in make_f0_v_sptk
    $FESTVOXDIR/src/general/smooth_f0 -o f0/$fname.f0 sptk_f0/$fname.f0 -otype ssff -lab lab/$fname.lab -silences $SILENCE  ....
where  SILENCE=awk '{print $1}' etc/silence
It looks fine but if you did not have a etc/silence` to start with, find_silence_name is invoked

(define (find_silence_name)
  (set! INST_LANG_VOX::clunits_prompting_stage t)
  (voice_INST_LANG_VOX_clunits)
  (set! silence (car (cadr (car (PhoneSet.description '(silences))))))
  (set! sfd (fopen "etc/silence" "w"))
  (format sfd "%s\n" silence)
  (fclose sfd)
)
and $SILENCE will be reduced to one silence name ..

By the way, despite its reassuring name "find_silence_name" is able to insidiously "destroy" a "etc/silence" if you are inside the directory of your voice and spits an error if you are not

When using setup_clunits to setup a voice project, there is no make_dur_model under folder bin, but build_clunits_voice need make_dur_model under folder bin.

Following the link "http://festvox.org/bsv/x3528.html", tried to built hindi tts from scratch on sample to 100 'hindi' wav files ['hindi_0001' - 'hindi_0102'] obtained from 'cmu_indic_hin_ab.tar.bz2'.

Every script works fine upto the following command--
./bin/do_clustergen parallel cluster etc/txt.done.data.train

getting lots of file not found & Segmentation fault (core dumped) Errors

example------------ final lines of above command:

Segmentation fault (core dumped)
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Attempt to access channel 1 of 0 channel track
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Segmentation fault (core dumped)
Segmentation fault (core dumped)
Dataset of 0 vectors of 64 parameters from: festival/feats/9r=_1.feats
Attempt to access channel 1 of 0 channel track
Dataset of 0 vectors of 64 parameters from: festival/feats/9r=_1.feats
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Attempt to access channel 1 of 0 channel track
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Segmentation fault (core dumped)
Dataset of 0 vectors of 64 parameters from: festival/feats/9r=_1.feats
Attempt to access channel 1 of 0 channel track
Dataset of 0 vectors of 64 parameters from: festival/feats/9r=_1.feats
Segmentation fault (core dumped)
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Attempt to access channel 1 of 0 channel track
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Segmentation fault (core dumped)
Segmentation fault (core dumped)
Dataset of 0 vectors of 64 parameters from: festival/feats/9r=_1.feats
Attempt to access channel 1 of 0 channel track
Dataset of 0 vectors of 64 parameters from: festival/feats/9r=_1.feats
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Attempt to access channel 1 of 0 channel track
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Segmentation fault (core dumped)
Dataset of 0 vectors of 64 parameters from: festival/feats/9r=_1.feats
Attempt to access channel 1 of 0 channel track
Dataset of 0 vectors of 64 parameters from: festival/feats/9r=_1.feats
Segmentation fault (core dumped)
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Attempt to access channel 1 of 0 channel track
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Segmentation fault (core dumped)
Dataset of 0 vectors of 64 parameters from: festival/feats/9r=_1.feats
Attempt to access channel 1 of 0 channel track
Dataset of 0 vectors of 64 parameters from: festival/feats/9r=_1.feats
Segmentation fault (core dumped)
Segmentation fault (core dumped)
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Attempt to access channel 1 of 0 channel track
Dataset of 0 vectors of 64 parameters from: festival/feats/9r_1.feats
Segmentation fault (core dumped)
Segmentation fault (core dumped)
Segmentation fault (core dumped)
Segmentation fault (core dumped)
Segmentation fault (core dumped)
Segmentation fault (core dumped)
Segmentation fault (core dumped)
Collect trees
SIOD ERROR: wrong type of argument to setcar
BACKTRACE:
0: (set-car! (car tree) vector_num)
1: (clustergen::dump_tree_vectors tree rawtrackfd)
2: (set! tree (clustergen::dump_tree_vectors tree rawtrackfd))
3: (f (car l2))
4: (cons (f (car l2)) r)
5: (set! r (cons (f (car l2)) r))
6: (while l2 (set! r (cons (f (car l2)) r)) (set! l2 (cdr l2)))
7: (mapcar
(lambda
(unit)
(...)
...)
unittypes)
8: (if
(consp cg:multimodel)
(mapcar
(...)
cg:multimodel)
...)
9: (begin
(set! cg:parallel_tree_build t)
(build_clustergen "etc/txt.done.data.train"))
closing a file left open: festival/trees/cmu_indic_ss_mcep.rawparams
closing a file left open: festival/trees/cmu_indic_ss_mcep.tree

I completed festival part successfully with 16kHz .wav files. But when I'm trying to use 44100Hz .wav files I'm getting errors. CAtually building prompts was okay. But the problem started in the labeling part. Only .sl files are created. .lab and .slehmm files are not created and it shows errors and labeling part stopped at the middle.
I need to know do we need to change or edit some files (codes) when we are using a sample rate other than 16kHz.
Please someone help me to overcome this issue.

Thanks in advance.

do_pm call ch_wave with the parameter F=16
but ch_wave -h says
-F Output sample rate in Hz. If this is different from the input sample rate, resampling will occur

and errors do occur (rateconv)

I was building indic voice using the steps given in the website and when I typed in
./bin/do_build build_prompts etc/txt.done.data It gave me error as below

SIOD ERROR: could not open file ./festvox/language_variant.scm
I got to know that it is because of the language selector in the festvox/indic_lexicon.scm

Am I wrong somewhere? or Is it still under development? Please let me know

I am currently running Festvox 2.8 with Festival 2.5

Hello, I'm building a voice in Brazilian Portuguese used "Grapheme-based in Grapheme-based Synthesizer, but in generating some test examples occurred:

./bin/do_clustergen cg_test tts tts etc/txt.done.data.test Resulting in SIOD ERROR: null-arg-to-last.

The database has 600 audios on average from 3 to 8 seconds long. All training commands and generation of label the data and prompts. and feature extraction are successful.

in /src/clustergen/fr/do_cga, you are supposed to

Build voice conversion model using CART prediction (and initially)

using a clustergen voice as a source

but you don't build with it your own files :
you use for example "$CLUSTERGENDIR/clustergen_build.scm"
that could not be the same as yours, particularly if work with a new language

this file is missing

Hi, I'm getting the following error when trying to build prompts.

free(): invalid next size (fast)
Aborted (core dumped)

Can someone tell me the reason and solution for this please.

Thank you