Build your own voice example for indic languages (Hindi TTS)

[prajwaljpj]

I am new to merlin. I have been trying to build a Hindi TTS system using my own data but end up failing at festival (Segmentation fault). I am currently using the latest versions of festival, speech_tools, festvox cloned from github. I also tried compiling all the tools with gcc-5 and gcc-7. Here is where it fails.

train_sentences.scm 
(utt.save (utt.synth (Utterance Text "इस अनाज की विशेषता यह थी कि कम लागत मे बहुत उपजता था." )) "/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//prompt-utt/10_1_1.utt")
(utt.save (utt.synth (Utterance Text "अमज़ान के सघन आदिम रेन फॉरेस्ट पेरु ब्रज़ील वेनेजुयेला इक्वेडोर बोलिविआ गुयेना." )) "/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//prompt-utt/10_1_10.utt")
(utt.save (utt.synth (Utterance Text "सारांश यह कि ईश्वर और महाराज एडवर्ड के बाद इस देश मे आपही का दरजा था." )) "/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//prompt-utt/10_1_100.utt")
(utt.save (utt.synth (Utterance Text "उससे कुछ नही होगा जेब मे रखकर चलता बनेगा." )) "/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//prompt-utt/10_1_101.utt")
(utt.save (utt.synth (Utterance Text "जी भरकर सुनाते हैं चाहे सामनेवाला उँघे या सो जाए." )) "/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//prompt-utt/10_1_102.utt")
(utt.save (utt.synth (Utterance Text "मंजुला ने गर्व के साथ कहा मै ऐसी आपत्तियों की परवाह नही करती." )) "/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//prompt-utt/10_1_103.utt")
(utt.save (utt.synth (Utterance Text "बोनू की आवाज़ मे दिल को छलनी कर देने वाली कात्रता थी." )) "/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//prompt-utt/10_1_104.utt") 
.
.
.

When I run Step 2 ./02_prepare_labels.sh

./02_prepare_labels.sh /home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/wavs/ /home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/txts/ /home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels/
Step 2: 
Preparing labels...
preparing full-contextual labels using Festival frontend...
creating a scheme file from text directory
generating utts from scheme file
/home/gnani/projects/merlin/egs/build_your_own_voice/s1/scripts/prepare_labels_from_txt.sh: line 53: 76192 Segmentation fault      (core dumped) ${FESTDIR}/bin/festival -b ${out_dir}/$scheme_file
converting festival utts to labels...
01_setup.sh 02_prepare_labels.sh 03_prepare_acoustic_features.sh 04_prepare_conf_files.sh 05_train_duration_model.sh 06_train_acoustic_model.sh 07_run_merlin.sh README.md conf database experiments run_demo_voice.sh scripts
gawk: fatal: cannot open file `/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//prompt-lab/tmp' for reading (No such file or directory)
gawk: fatal: cannot open file `/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//prompt-lab/tmp' for reading (No such file or directory)
normalizing label files for merlin...
10_1_1.lab
Traceback (most recent call last):
  File "/home/gnani/projects/merlin/misc/scripts/frontend/utils/normalize_lab_for_merlin.py", line 134, in <module>
    normalize_label_files(in_lab_file, out_lab_file, label_style, write_time_stamps)
  File "/home/gnani/projects/merlin/misc/scripts/frontend/utils/normalize_lab_for_merlin.py", line 35, in normalize_label_files
    in_f = open(in_lab_file,'r')
FileNotFoundError: [Errno 2] No such file or directory: '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//prompt-lab/full/10_1_1.lab'
forced-alignment using HTK tools...
---preparing enverionment
---checking data
---extracting features
Mon Jul 13 09:29:14 2020
---feature_normalisation
Mon Jul 13 09:29:14 2020
---making proto
Mon Jul 13 09:29:14 2020
---training HMM models
  ERROR [+7220]  NewHMMScan: cannot find any physical HMMs to scan
 FATAL ERROR - Terminating program /home/gnani/projects/merlin/tools/bin/htk/HERest
Traceback (most recent call last):
  File "/home/gnani/projects/merlin/misc/scripts/alignment/state_align/forced_alignment.py", line 381, in <module>
    aligner.train_hmm(7, 32)
  File "/home/gnani/projects/merlin/misc/scripts/alignment/state_align/forced_alignment.py", line 285, in train_hmm
    stdout=PIPE)
  File "/usr/lib/python3.6/subprocess.py", line 311, in check_call
    raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['/home/gnani/projects/merlin/tools/bin/htk/HERest', '-C', '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels/config/cfg', '-S', '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels/config/train.scp', '-I', '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels/config/mono_phone.mlf', '-M', '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels/model/hmm_mix_1_iter_1', '-H', '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels/model/hmm0/macros', '-H', '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels/model/hmm0/hmmdefs', '-t', '250.0', '150.0', '2000.0', '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels/mono_phone.list']' returned non-zero exit status 52.
ls: cannot access '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//label_state_align': No such file or directory
Force-alignment unsucessful!! Please check /home/gnani/projects/merlin/misc/scripts/alignment/state_align/forced_alignment.py
Copying labels to duration and acoustic data directories...
cp: cannot stat '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//label_state_align': No such file or directory
cp: cannot stat '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//label_state_align': No such file or directory
ls: cannot access '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//label_state_align': No such file or directory
ls: cannot access '/home/gnani/merlin/egs/build_your_own_voice/s1/female_hin/labels//label_state_align': No such file or directory
done...!

Any help would be greatly appreciated! Thank you!