I am defining the new voice (Punjabi) in festival using diphone concatenation approach. I have created all the necessary files. Firstly I am trying toy example for one word i.e. "ksr" ("ਕਸਰ" in punjabi ) I have added phones and their features in phoneset.scm file next, I added lexicon entry in lexicon.scm file. i have also defined letter to sound rule by hand. also made lex.out text file with word "ਕਸਰ" and its syllablic structure. I have also set durations of phones in dur.data file
I have recorded the nonsense words wave files for diphones k-s (ਕ-ਸ) and s-r (ਸ-ਰ) using the audacity recororder and manually identified the start and end of the diphone also created file text.done.data with the following entries
( pa_0029 "pau t aa k - s aa t aa pau" ("k-s") )
( pa_1034 "pau t aa s - r aa t aa pau" ("s-r") )
and created .est index file with entries and like
EST_File index
DataType ascii
NumEnteries 2
IndexName pn_diphone
EST_Header_End
k-s pa_0029 1.173 1.267 1.361
s-r pa_1034 1.142 1.229 1.316
after that i extract pitchmarks etc with the following commands make_pm_wav, make_pm_fix make_pm_lpc like this
pansy@ubuntu:~/fest/net_pa_pn_diphone$ bin/make_pm_wave wav/*.wav
pa_0029 PM_WAVE
pa_1034 PM_WAVE
after running command festival festvox/net_pa_pn_diphone.scm "(voice_net_pa_pn_diphone)" it enters the festival but before that it says
EST_Option: No value set for NumEntries
I change the voice
festival> (voice_net_pa_pn_diphone)
net_pa_pn_diphone
when i run say phones command in festival it gives the following error:
festival> (SayPhones '(k s r))
US DB: can't find diphone k-s and even default diphone (pau-pau) doesn't exist
-=-=-=-=-=- EST Error -=-=-=-=-=-
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
What is the solution for this problem? I am using festival 2.4.
