Bugzilla Bug 211214

cmu voices don't work:
(set! voice_default 'voice_cmu_us_awb_arctic_hts)

default voice and "ked" works:
(set! voice_default 'voice_ked_diphone)

Reproducible: Always

Steps to Reproduce:
# echo "Hello. I'm the Apple Powerbook G4 running Gentoo Linux." |festival
--tts 
 SIOD ERROR: module hts_engine required, but not compiled in this installation 

 closing a file left open:
/usr/share/festival/voices/us/cmu_us_slt_arctic_hts/festvox/cmu_us_slt_arctic_hts.scm 
 closing a file left open: /usr/share/festival/init.scm 
 festival: fatal error exiting.


Expected Results:  
Was working with 1.95_beta-r4

This bug is because the HTS module is not compiled. The reason is the ebuild
line:

# disable the multisyn modules
sed -i -e "s:\(ALSO_INCLUDE.*=.*MultiSyn\):# \1:" "${S}"/config/config.in

which comments out all the line, including HTS module, not only the unstable
MultiSyn.

I just looked this over, and the issue is that the klunits, hts_engine
and MultiSyn modules are all considered experimental by upstream, and
I did not take that into account in the previous ebuilds.

So, there are two options.  Let me know if you would prefer that I
remove the voices or bring back the experimental modules.

Thanks,

William

Since I'm using the experimental in the previous version without problems I
would say 'bring it back'. People can switch to standard voices if they like.

(In reply to comment #2)

> So, there are two options.  Let me know if you would prefer that I
> remove the voices or bring back the experimental modules.

Make them a USE option?

I would also like it to be brought back, or else made into some kind of USE
option.

Please, bring back the hts option.  The default voice is just too... 
computer-y.  I used to use the female one, as it sounded the most
human-sounding, computer voice.  It requires the hts engine, though.

All,

I have a local festival ebuild with the extra engines enabled -- I have
deleted the lines mentioned in comment #1.  I get the following:

william@linux1 ~ $ cat .festivalrc
(set! voice_default 'voice_cmu_us_awb_arctic_hts)
;; (Parameter.set 'Audio_Command "aplay -q -c 1 -t raw -f s16 -r $SR $FILE")
;; (Parameter.set 'Audio_Method 'Audio_Command)
william@linux1 ~ $ echo "gentoo can speak." | festival --tts
SIOD ERROR: unbound variable : f2b_f0_lr_start
closing a file left open: /usr/share/festival/init.scm
festival: fatal error exiting.
william@linux1 ~ $ 

I do not know the internals of festival all that well, so can someone provide
information about this issue?

Thanks,

William

William, according to the wiki (http://gentoo-wiki.com/HOWTO_speechd#HTS) the
problem you had above can be easily fixed. However, how did you build it? I've
tried commenting out that line, but I get an error with MultiSyn when
compiling...

I also tried commenting out the following:
sed -i -e "s:\(ALSO_INCLUDE.*=.*MultiSyn\):# \1:" "${S}"/config/config.in


However, the build errored out with:
making dependencies -- hash_s_itemlistp_t.cc hash_s_itemlistp_t.cc:54:38:
error: ../base_class/EST_THash.cc: No such file or directory

Would this have something to do with?:
sed -i -e "s:\(EST=\).*:\1/usr/share/speech-tools:" "${S}"/config/confi$
sed -i -e "s:\$(EST)/lib:/usr/$(get_libdir):" "${S}"/config/project.mak

I got it working by simply pausing portage right after it unpacked, editing out
MultiSyn on line 58 of config/config.in, and letting it continue. After that,
it worked fine with all the arctic and nitech voices I downloaded.

The speech engines are added back now, excluding the multisyn engine.

Several of the CMU voices have issues in their scheme files, so I am
assigning this to one of our developers who is familiar with scheme so
he can take a look at them.

I tried to load the arctic voices and ran into the error described in this
posting on the hts-users list.  Applied their changes to all the arctic voices
and they load now and say the introduction.

http://hts.sp.nitech.ac.jp/hts-users/spool/2007/msg00068.html

I am having trouble getting duration changes to apply to the arctic voices but
that might be a "missing feature".

Created an attachment (id=180019) [details]
files/festival-1.96_beta-hts-voices.patch

thanks for the above link, I've created a patch to fix the cmu voice files for
1.96, see attached

all you need to do
1. create a copy of the ebuild in an overlay directory e.g.
festival-1.96_beta-r1.ebuild
2. add epatch "${FILESDIR}"/${P}-hts-voices.patch under the unpack section and
save the attached file into the files directory
3. run ebuild digest on the new ebuild
4. and emerge

Created an attachment (id=180022) [details]
festival-voices-us-nitech-hts-2.1.ebuild

an ebuild to add in the hts nitech voices

Created an attachment (id=180023) [details]
festival-1.96_beta-r1.ebuild

the current ebuild I'm using for festival at the moment to add in the patch

I am re-assigning this back to myself since pchrist is away.

The issue with the CMU voices not working has been fixed in -r1.

Since the us-nitech voices are a separate package and not part of the
actual festival distribution, please file a separate bug requesting a new
ebuild.

Thanks much.

William