making the .flitevox voices from source?

[thebunnyrules]

Hi guys, great great project. These voices are really amazing.

I've built flite 2.1 from source (probably one of the smoothest builds I've had in Linux) but I noticed that the .flitevox voices need to be downloaded instead of built from source?

I was wondering what the procedure for building these voices is and where I can find the source code?

[saikrishnarallabandi]

Hi,

Sorry for the delay in response. flitevox voices can be built. I am assuming that you mean from a festival voice directory when you say source.

The following are set of commands to build a flite voice from existing festival voice directory:

$FLITEDIR/tools/setup_flite ./bin/build_flite cd flite make

If you also want to dump the voice, make voicedump make add_feats

An example script which has these is build_cg_rfs_voice (It should be in the bin directory)

Hope this helps!

[thebunnyrules]

@saikrishnarallabandi thanks so much for the reply. sorry I didn't answer before now. I must have missed github's email notification.

I've tried to follow your directions but I'm a little new at this and there's a couple of points I'm confused about.

$FLITEDIR/tools/setup_flite

By FLITEDIR, do you mean the flite2.1 source code master dir that I clone from github? I'm asking because I compiled and installed flite2.1 but there was no tools folder in the final install prefix. On the other hand, there is a tools folder in the source code directory with a setup_flite script and there is a bin folder with a build_flite binary that appears after the make step in compilation. Are you referring to that in your ./bin/build_flite step? I just want to make sure that my assumptions are correct here and that you're not referring to something else.

am assuming that you mean from a festival voice directory when you say source.

I found a couple of sources: for this build I'm trying to build cmu_us_fem (the us english with german accent). I found a source for it in the festival arctic section found in: http://festvox.org/cmu_arctic/packed/cmu_us_fem_arctic.tar.bz2 (the one with /etc/txt.done.data and /wav/arctic_a000x.wav files). I also found http://festvox.org/packed/festival/2.5/voices/festvox_cmu_us_fem_cg.tar.gz.

I tried your procedure with both of these sources, setting "export FLITEDIR=git0cloned-flite2.1-sourcefolder-with-compiled-binaries" and "export ESTDIR=cloned-speech_tools-sourcefolder-with-compiled-binaries"

When I used $FLITEDIR/tools/setup_flite on extracted festvox_cmu_us_fem_cg.tar.gz folder, script said that it was missing the /wav dir. When I tried it on cmu_us_fem_arctic.tar.bz2's extracted folder, I got this error message:

Can't find etc/voice.defs file don't know what voice to convert Setup for conversion of to flite FAILED

Do I have the wrong source files? Or is there a step I'm missing or did wrong?

[saikrishnarallabandi]

Hi,

Saw this just now. sorry. Lets start with this directory you downloaded: http://festvox.org/cmu_arctic/packed/cmu_us_fem_arctic.tar.bz2

Once you extract, you can create the voice specific variables by running the following:

Setup Voice directory

$FESTVOXDIR/src/clustergen/setup_cg cmu us fem

Ignore the warning and let it populate the folders. Now, you should be able to see the binaries in bin/

Running those shoud suffice. Let me know if there is some issue .

[thebunnyrules]

Hello,

I feel like I'm really close but it's still giving me some errors close to the last step. so this is what I have been able to put together from your instructions so far:

Short version: build flite, build speech tools, build festival (this wasn't in your instructions but it solved an error I was getting during the ./bin/build_flite step), build festvox, extract cmu_us_fem_arctic.tar.bz2 and go to extracted folder, $FESTVOXDIR/src/clustergen/setup_cg cmu us fem, $FLITEDIR/tools/setup_flite, ./bin/build_flite, cd flite, make

Long version:

#Get source for flite, festvox, speech tools, festival and a voice from here http://festvox.org/cmu_arctic/packed (for this example: http://festvox.org/cmu_arctic/packed/cmu_us_fem_arctic.tar.bz2)

#the paths will vary on your PC. replace the paths below with the ones you're using.

#goto flite-master dir (flite source code) and
export FLITEDIR=/home/ubuntu/Desktop/festival-tests/flite
cd $FLITEDIR; ./configure; make

#goto speech_tools master dir and
export ESTDIR=/home/ubuntu/Desktop/festival-tests/speech_tools
cd $ESTDIR; ./configure; make

#goto festival dir (seems to be needed during voice build)
export FESTIVALDIR=/home/ubuntu/Desktop/festival-tests/festival
cd $FESTIVALDIR; ./configure; make; make default_voices

#goto festvox master dir and
export FESTVOXDIR=/home/ubuntu/Desktop/festival-tests/festvox
cd $FESTVOXDIR; ./configure; make

#goto downloaded voice, extract, goto dir:
cd /home/ubuntu/Desktop/festival-tests/cmu_us_fem_arctic
$FESTVOXDIR/src/clustergen/setup_cg cmu us fem
$FLITEDIR/tools/setup_flite
./bin/build_flite
cd flite
make

#If you also want to dump the voice,
make voicedump
make add_feats

Errors:

During the ./bin/build_flite step (for some reason it's sliping a // into the paths it's looking into)

ubuntu@ubuntu:~/Desktop/festival-tests/cmu_us_fem_arctic$ ./bin/build_flite
cg_convert: finding parameter ranges
Error reading ESPS file festival/trees/cmu_us_fem_mcep*.params
cat: etc/f0.params: No such file or directory
Error reading ESPS file /home/ubuntu/Desktop/festival-tests/cmu_us_fem_arctic//festival/trees/cmu_us_fem_mcep.params
Cannot load track: /home/ubuntu/Desktop/festival-tests/cmu_us_fem_arctic//festival/trees/cmu_us_fem_mcep.params
SIOD ERROR: could not open file /home/ubuntu/Desktop/festival-tests/cmu_us_fem_arctic//festival/trees/cmu_us_fem_mcep.tree
flite_build cg complete.  You can compile the generated voice by
   cd flite; make

during the: cd flite; make step (probably because the preceding step had failed):

cd flite; make
ls: cannot access 'cmu_us_fem_cg_*_f0_trees.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_mcep_trees.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_params.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_durmodel.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_f0_trees.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_mcep_trees.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_params.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_durmodel.c': No such file or directory
make: *** No rule to make target 'paramfiles.mak'.  Stop.

[saikrishnarallabandi]

Oh sorry. Can you build the festvox voice and then get to flite voice? I assumed you had that voice already.

Here are the steps to build: ( after putting txt.done.data in the folder 'etc' and copying the wavefiles to 'wav'

Some text processing

./bin/do_build build_prompts etc/txt.done.data ./bin/do_build label etc/txt.done.data ./bin/do_clustergen parallel build_utts etc/txt.done.data ./bin/do_clustergen generate_statenames ./bin/do_clustergen generate_filters

Extract acoustic features

./bin/do_clustergen parallel f0_v_sptk ./bin/do_clustergen parallel mcep_sptk ./bin/do_clustergen parallel str_sptk ./bin/do_clustergen parallel combine_coeffs_me cp festvox/clustergen.scm festvox/clustergen.scm.xxx cat festvox/clustergen.scm.xxx | sed 's/mixed_excitation nil/mixed_excitation t/' | cat >festvox/clustergen.scm

Make train test split

./bin/traintest etc/txt.done.data

Train a model

./bin/do_clustergen parallel cluster etc/txt.done.data.train ./bin/do_clustergen dur etc/txt.done.data.train

Now do the flite.

It should take an hour to finish this.

[thebunnyrules]

Sure, no problem. I'll give that a go. Should I do these steps before before I do the step with: $FESTVOXDIR/src/clustergen/setup_cg cmu us fem or after?

[saikrishnarallabandi]

After.

$FESTVOXDIR/src/clustergen/setup_cg cmu us fem

sets up the folder structure.

[thebunnyrules]

Hi Sai Krishna,

I followed your instructions but still got this error at the end:

cd flite; make #gives these errors:
ls: cannot access 'cmu_us_fem_cg_*_f0_trees.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_mcep_trees.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_params.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_durmodel.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_f0_trees.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_mcep_trees.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_params.c': No such file or directory
ls: cannot access 'cmu_us_fem_cg_*_durmodel.c': No such file or directory

When I downloaded this file: http://www.festvox.org/packed/festival/2.5/voices/festvox_cmu_us_fem_cg.tar.gz and put the contents of contents of /festival/lib/voices/us/cmu_us_fem_cg into the directory where I was building the voices and re-did:

./bin/build_flite; cd flite; make

it succeeded. This is great but festvox_cmu_us_fem_cg.tar.gz is a packed version of the voice with a bunch of param binaries in it. I'm try to do everything from source. Do you know how I can go about generating these files?

files that I have not been able to generate so far: ``` festvox/cmu_us_fem_dur.scm dur_rf_models/mlist dur_rf_models/dur_14/cmu_us_fem_durdata_cg.scm dur_rf_models/dur_08/cmu_us_fem_durdata_cg.scm dur_rf_models/dur_01/cmu_us_fem_durdata_cg.scm rf_models/mlistf0 rf_models/trees_14/cmu_us_fem_f0.tree rf_models/trees_13/cmu_us_fem_f0.tree rf_models/trees_10/cmu_us_fem_f0.tree rf_models/mlist rf_models/trees_09/cmu_us_fem_mcep.params rf_models/trees_09/cmu_us_fem_mcep.tree rf_models/trees_09/cmu_us_fem_f0.tree rf_models/trees_06/cmu_us_fem_mcep.params rf_models/trees_06/cmu_us_fem_mcep.tree rf_models/trees_06/cmu_us_fem_f0.tree rf_models/trees_19/cmu_us_fem_mcep.params rf_models/trees_19/cmu_us_fem_mcep.tree rf_models/trees_19/cmu_us_fem_f0.tree festival/trees/cmu_us_fem_mcep.params festival/trees/cmu_us_fem_mcep.tree festival/trees/cmu_us_fem_f0.tree ``` This is a list of the files I was able to generate in the absence of festvox_cmu_us_fem_cg.tar.gz : https://pastebin.com/ysgZk9pg

[saikrishnarallabandi]

This uses Random Forests.

./bin/build_cg_rfs_voice should build those dependencies.

( It will run for around 12-15 hours depending on the config)

[thebunnyrules]

Hey, before we continue, I just want to say thank you. You've really gone out of your way to help me and I really appreciate that.

At very bottom of this comment is my updated recipe for building a voice. At what point should I use _./bin/build_cg_rfsvoice in these steps? I've tried to do it after the _./bin/doclustergen dur etc/txt.done.data.train step but it crashes after an hour. I'm going to play around with my setup and see if I can fix the errors on my own. I'll get back to you later on when I have a clearer picture.

Updated Voice Building Recipe

(skipping the tool building steps for the sake of brevity)

#Do text processing
./bin/do_build build_prompts etc/txt.done.data
./bin/do_build label etc/txt.done.data
./bin/do_clustergen parallel build_utts etc/txt.done.data
./bin/do_clustergen generate_statenames
./bin/do_clustergen generate_filters

#Extract acoustic features
./bin/do_clustergen parallel f0_v_sptk
./bin/do_clustergen parallel mcep_sptk
./bin/do_clustergen parallel str_sptk
./bin/do_clustergen parallel combine_coeffs_me
cp festvox/clustergen.scm festvox/clustergen.scm.xxx;cat festvox/clustergen.scm.xxx | sed 's/mixed_excitation nil/mixed_excitation t/' > festvox/clustergen.scm

#Make train test split
./bin/traintest etc/txt.done.data

#Train a model
./bin/do_clustergen parallel cluster etc/txt.done.data.train
./bin/do_clustergen dur etc/txt.done.data.train

#Now do the flite.
$FLITEDIR/tools/setup_flite
./bin/build_flite
cd flite; make

#If you also want to dump the voice,
make voicedump
make add_feats

[thebunnyrules]

Sai Krishna, in your first post, you said to that after I succeed in making a voice dump (make voicedump), I can use the command "make add_feats" to add feats. As, I had stated above, using the CMU generated files found in festvox_cmu_us_fem_cg.tar.gz, I succeed in generating a voice but I forgot to mention that when I tried to do the "make add_feats" command after generating the voicedump, I got this error: "make: *** No rule to make target add_feats. Stop." I've tried using rgrep to search all the source and config files in festival,festvox,flite and speech_tools for the add_feats command and I couldn't find it. Maybe it's been renamed to something new? Do you have any suggestions?

[saikrishnarallabandi]

Ok. Can you clear the folders and build from scratch?

1) make sure folder structure is present. $festvoxdir/src/clustergen/setup_cg Cmu us fem 2) copy the txt.done.data and wavefiles 3) run ./bin/build_cg_rfs_voice.

• sent from my Android

On Wednesday, May 2, 2018, thebunnyrules notifications@github.com wrote:

Sai Krishna, in your first post, you said to that after I succeed in making a voice dump (make voicedump), I can use the command "make add_feats" to add feats. As, I had stated above, using the CMU generated files found in festvox_cmu_us_fem_cg.tar.gz, I succeed in generating a voice but I forgot to mention that when I tried to do the "make add_feats" command after generating the voicedump, I got this error: "make: *** No rule to make target add_feats. Stop." I've tried using rgrep to search all the source and config files in festival,festvox,flite and speech_tools for the add_feats command and I couldn't find it. Maybe it's been renamed to something new? Do you have any suggestions?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/festvox/flite/issues/6#issuecomment-385876071, or mute the thread https://github.com/notifications/unsubscribe-auth/AV3IMcizBrjheNjAfYcLJIN-vYdV9W4Hks5tuVBHgaJpZM4RY8qw .

[thebunnyrules]

I've aleardy done this several times (rebuilt voice and tools from sctach). At no point is add_feats available.

[Pranav-Jain]

Hi, I am running into error in the final step to create voices using flite and I hope you would help. I have successfully trained the model and then followed the following steps. $FLITEDIR/tools/setup_flite ./bin/build_flite cd flite; make The final make command works great.

But finally, when I try to run the binary created by the "make" to create voices from a given text, I run into a segmentation fault.

[Brajkishore-ai]

./bin/do_build build_prompts etc/txt.done.data

Hi, While running the do_build I am getting the belo error.

SIOD ERROR: could not open file /projects/festival/lib/dicts/cmu/cmulex.scm closing a file left open: ./festvox/indic_lexicon.scm closing a file left open: ./festvox/cmu_indic_guj_lexicon.scm closing a file left open: festvox/cmu_indic_guj_clunits.scm closing a file left open: festvox/build_clunits.scm

I am not able to find the cmulex.scm file due to which the error is occuring. I am running this file from the voice folder that I have downloaded and extracted. Can you please provide a solution to this or any suggestion to solve this issue. Thank you

[awbcmu]

You need to install the cmulex, in the festival/ source directory

make default_voices

which simply runs the script

./src/scripts/default_voices.sh

which installs a default voice and the default lexicons

Alan

On Fri, Sep 29, 2023 at 3:36 AM Brajkishore @.***> wrote:

./bin/do_build build_prompts etc/txt.done.data

Hi, While running the do_build I am getting the belo error.

SIOD ERROR: could not open file /projects/festival/lib/dicts/cmu/cmulex.scm closing a file left open: ./festvox/indic_lexicon.scm closing a file left open: ./festvox/cmu_indic_guj_lexicon.scm closing a file left open: festvox/cmu_indic_guj_clunits.scm closing a file left open: festvox/build_clunits.scm

I am not able to find the cmulex.scm file due to which the error is occuring. I am running this file from the voice folder that I have downloaded and extracted. Can you please provide a solution to this or any suggestion to solve this issue. Thank you

— Reply to this email directly, view it on GitHub https://github.com/festvox/flite/issues/6#issuecomment-1740747312, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAOEXNHGACHS3CHU6G7WB63X42XCXANCNFSM4ELDZKYA . You are receiving this because you are subscribed to this thread.Message ID: @.***>