[prev in list] [next in list] [prev in thread] [next in thread] 

List:       festlang-talk
Subject:    [festival-talk] Voice (slt_arctic_hts) sounds different within
From:       joe () skyrush ! com (Joe Peterson)
Date:       2005-08-17 13:27:14
Message-ID: 43033B32.3030309 () skyrush ! com
[Download RAW message or body]

message from Joe Peterson <joe at skyrush.com> to festival-talk
= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
Here's something interesting...

I went to the HTS web site (http://hts.ics.nitech.ac.jp/) and downloaded
the voice from there.  It was rooted in the "lib" dir.  I moved the
cmu_us_slt_arctic_hts aside and replaced it with the HTS one.  Good news
is that it sounds like the demo web site!  Bad news is that is Seg
Faults at "line 1" of something (see below).  I'm not as surprized about
this as I am that it still manages to work.  Is there a simple fix to
eliminate the Seg Fault?  BTW, I'm pretty new to Festival...

Festival Speech Synthesis System 1.96:beta July 2004
Copyright (C) University of Edinburgh, 1996-2004. All rights reserved.
For details type `(festival_warranty)'
festival> (voice_cmu_us_slt_arctic_hts)
cmu_us_slt_arctic_hts
festival> (SayText "Yo, dude")
sh: line 1: 15673 Segmentation fault
/home/joe/festival/festival/lib/voices/us/cmu_us_slt_arctic_hts/hts//hts_engine
-dm
/home/joe/festival/festival/lib/voices/us/cmu_us_slt_arctic_hts/hts//mcep_dyn.win
-dm
/home/joe/festival/festival/lib/voices/us/cmu_us_slt_arctic_hts/hts//mcep_acc.win
-df
/home/joe/festival/festival/lib/voices/us/cmu_us_slt_arctic_hts/hts//lf0_dyn.win
-df
/home/joe/festival/festival/lib/voices/us/cmu_us_slt_arctic_hts/hts//lf0_acc.win
-td
/home/joe/festival/festival/lib/voices/us/cmu_us_slt_arctic_hts/hts//tree-dur.inf
-tm
/home/joe/festival/festival/lib/voices/us/cmu_us_slt_arctic_hts/hts//tree-mcep.inf
-tf
/home/joe/festival/festival/lib/voices/us/cmu_us_slt_arctic_hts/hts//tree-lf0.inf
-md
/home/joe/festival/festival/lib/voices/us/cmu_us_slt_arctic_hts/hts//duration.pdf
-mm
/home/joe/festival/festival/lib/voices/us/cmu_us_slt_arctic_hts/hts//mcep.pdf
-mf
/home/joe/festival/festival/lib/voices/us/cmu_us_slt_arctic_hts/hts//lf0.pdf
-a 0.420000 -r 0.000000 -fs 1.000000 -fm 0.000000 -om
/tmp/est_15667_00000/tmp.mcep -of /tmp/est_15667_00000/tmp.f0 -or
/tmp/est_15667_00000/tmp.raw -u 0.500000 -l 0.000000 -b 0.300000
/tmp/est_15667_00000/tmp.lab >/dev/null
#<Utterance 0xb72194f8>


-Joe



Alan W Black wrote:
>  Joe Peterson writes on 16 August 2005:
>  > message from Joe Peterson <joe at skyrush.com> to festival-talk
>  > = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
>  > Sorry if this has been discussed, but I cannot find the answer...
>  > 
>  > I tried the demo on:
>  > 
>  > 	http://www.festvox.org/voicedemos.html
>  >  	Voice used: slt_arctic_hts (American Female)
>  > 	Text: "searching for truth"
>  > 
>  > It sounds great!
>  > 
>  > Now, when I run festival (v. 1.95), and type:
>  > 
>  > 	festival> (voice_cmu_us_slt_arctic_hts)
>  > 	cmu_us_slt_arctic_hts
>  > 	festival> (SayText "searching for truth")
>  > 	#<Utterance 0xb716c838>
>  > 
>  > I get quite a different result.  The pronounciation is quite different.
>  >  For example, the "for" sound like "far," and "truth" sounds wavery.
> 
> The on-line demo uses a non-intgerated version of the voice. (Separate
> binaries are need for Festival and hts).  1.96 has a closer feel
> to the two-binaries one, but they are still different (and were
> trained at different times).  We will update this before 2.0.  but
> in the mean time you probably want 1.96
> 
> http://www.speech.cs.cmu.edu/awb/fftest/speech_tools-1.2.96-beta.tar.gz
> http://www.speech.cs.cmu.edu/awb/fftest/festival-1.96-beta.tar.gz
> http://www.speech.cs.cmu.edu/awb/fftest/festlex_CMU.tar.gz
> http://www.speech.cs.cmu.edu/awb/fftest/festvox_cmu_us_awb_arctic_hts.tar.gz
> http://www.speech.cs.cmu.edu/awb/fftest/festvox_cmu_us_bdl_arctic_hts.tar.gz
> http://www.speech.cs.cmu.edu/awb/fftest/festvox_cmu_us_jmk_arctic_hts.tar.gz
> http://www.speech.cs.cmu.edu/awb/fftest/festvox_cmu_us_slt_arctic_hts.tar.gz
> 
> 
>  > I must not be setting up some parameters that the web demo uses, but I
>  > cannot find any info on this.  If anyone knows, please let me know how
>  > to make festival sound more like the demo.
>  > 
>  > 	Thanks, Joe
>  > = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
>  > =    University of Edinburgh's Festival Speech Synthesis System       =
>  > = http://festvox.org/festival      Sent Via festival-talk at festvox.org =
>  > =                           To unsubscribe mail majordomo at festvox.org =
>  > = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
>  > 
> 
= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
=    University of Edinburgh's Festival Speech Synthesis System       =
= http://festvox.org/festival      Sent Via festival-talk at festvox.org =
=                           To unsubscribe mail majordomo at festvox.org =
= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =



[prev in list] [next in list] [prev in thread] [next in thread]