[mary-users] No sound while using our own Turkish voice
Marcela Charfuelan
Marcela.Charfuelan at dfki.de
Thu May 5 13:40:59 CEST 2011
Hello,
I am afraid the only solution right now, would be to downsample your DB
to 16K, I found that the vocoder has hard coded that sampling rate
(bad.... ) , so even if you change the sampling rate in the
configuration file, you will not be able to get proper sound.
sorry about that. But also, the filters we use for mixed excitation
were designed for 16K, so if you are using 44K maybe you need to tune a
bit those filters.
I am updating right now the hts_engine to version 1.04 so this problem
will be solved in the next release. Thanks for reporting it!
Regards,
Marcela.
On 05/04/2011 10:30 PM, Emre Çelikten wrote:
> Hello,
>
> Thanks very much for your help. I have found the problem. My library
> was in 44100 Hz sample rate. Apparently 44100 is not supported:
>
> 2011-05-04 22:57:28,207 [RH 0] INFO marytts.Vocoder length in
> samples=143040
> 2011-05-04 22:57:28,212 [RH 0] INFO marytts.Synthesis Audio
> format conversion required for voice iyte_betul
> 2011-05-04 22:57:28,219 [RH 0] ERROR marytts.RH 0 Processing failed.
> java.lang.Exception: Module Synthesis: Problem processing the data.
> at marytts.server.Request.processOneChunk(Request.java:556)
> at
> marytts.server.Request.processOrLookupOneChunk(Request.java:394)
> at marytts.server.Request.process(Request.java:341)
> at
> marytts.server.http.SynthesisRequestHandler$1.run(SynthesisRequestHandler.java:261)
> Caused by: javax.sound.sampled.UnsupportedAudioFileException:
> Conversion from audio format PCM_SIGNED 16000.0 Hz, 16 bit, mono,
> 2 bytes/frame, little-endian to requested audio format PCM_SIGNED
> 44100.0 Hz, 16 bit, mono, 2 bytes/frame, little-endian not supported.
> Unsupported conversion: PCM_SIGNED 44100.0 Hz, 16 bit, mono, 2
> bytes/frame, little-endian from PCM_SIGNED 16000.0 Hz, 16 bit,
> mono, 2 bytes/frame, little-endian
> at
> marytts.modules.Synthesis.synthesizeOneSection(Synthesis.java:286)
> at marytts.modules.Synthesis.process(Synthesis.java:235)
> at marytts.server.Request.processOneChunk(Request.java:554)
> ... 3 more
>
>
> When I change the sample rate from conf folder I am able to receive
> sound but it is pitched down. What do you suggest? Should I downsample
> my library to 16000 Hz and rebuild the voice from scratch?
>
> I think server getting stuck in an infinite loop is a bug though. This
> is what I get in the console after waiting for a few minutes:
>
> Exception in thread "I/O dispatcher 5" Exception in thread "I/O
> dispatcher 3" java.lang.OutOfMemoryError: Java heap space
> at java.util.HashMap.newKeyIterator(HashMap.java:840)
> at java.util.HashMap$KeySet.iterator(HashMap.java:874)
> at java.util.HashSet.iterator(HashSet.java:153)
> at
> sun.nio.ch.SelectorImpl.processDeregisterQueue(SelectorImpl.java:127)
> at sun.nio.ch.EPollSelectorImpl.doSelect(EPollSelectorImpl.java:69)
> at sun.nio.ch.SelectorImpl.lockAndDoSelect(SelectorImpl.java:69)
> at sun.nio.ch.SelectorImpl.select(SelectorImpl.java:80)
> at
> org.apache.http.impl.nio.reactor.AbstractIOReactor.execute(AbstractIOReactor.java:235)
> Exception in thread "main" java.lang.OutOfMemoryError: GC overhead
> limit exceeded
> Exception in thread "I/O dispatcher 2" Exception in thread "I/O
> dispatcher 4" java.lang.OutOfMemoryError: GC overhead limit exceeded
>
> I think it also blocks the sound driver in Linux when it gets stuck.
>
> (PS: Sorry for sending the previous response directly to you.)
>
> Best,
>
> Emre
>
>
> On 05/04/2011 03:54 PM, Marcela Charfuelan wrote:
>> Hello,
>>
>> Well, difficult to say what is wrong, one idea that might help would
>> be to get DEBUG info from the server and check if there is any
>> problem when loading the voice. You can get debug information from
>> the server adding:
>>
>> -Dlog4j.logger.marytts=DEBUG,stderr
>>
>> on the java call in your: MARY_TTS/bin/maryserver
>>
>> Regards,
>> Marcela.
>>
>> On 05/04/2011 12:58 PM, Emre Çelikten wrote:
>>> Hello,
>>>
>>> Thanks for your answer.
>>>
>>> I have 10 wav files in that folder (along with raw and trace). They
>>> sound good except some rare clipping and labeling accuracy problems.
>>>
>>> I am using MARY 3.4.0 and the HTS distribution that is included in
>>> it. hts_engine is at v1.01.
>>>
>>> Best,
>>>
>>> Emre
>>>
>>>
>>> On 05/04/2011 01:46 PM, Marcela Charfuelan wrote:
>>>> Hello Emre,
>>>>
>>>> - you can check if the trained models actually create any sound
>>>> using the original hts_engine. You can see that in the directory
>>>> where you created your voice, there must be some wav files
>>>> generated (synthesised) by the original hts_engine, you can find
>>>> them in:
>>>> ../your_voice/hts/gen/qst001/ver1/hts_engine/*.wav
>>>> check if those files are generated and if they sound ok.
>>>>
>>>> - which versions of HTS and MARY are you using?
>>>>
>>>> Regards,
>>>> Marcela.
>>>>
>>>>
>>>>
>>>> On 05/04/2011 12:20 PM, Emre Çelikten wrote:
>>>>> Hello,
>>>>>
>>>>> I am sorry if something similar has been asked before.
>>>>>
>>>>> I have run across a problem while using MARY. I have built a
>>>>> Turkish HMM voice by following the tutorial on the wiki. Every
>>>>> step in Voice Import Tools is completed successfully. Installing
>>>>> the voice using the component installer is successful as well.
>>>>>
>>>>> I cannot hear any sound when I try to synthesize audio using the
>>>>> voice that I have built. There are no error messages from the
>>>>> server or client but a Java process spawned from the server seems
>>>>> to be using 100% of a processor. I have waited for a bit hoping
>>>>> that the computation will finish, to no avail.
>>>>>
>>>>> Other voices work fine including dfki-ot-hsmm. Also, another voice
>>>>> that I have created (for English) using CMU ARCTIC AWBdatabase
>>>>> works fine as well.
>>>>>
>>>>> Do you have any suggestions?
>>>>>
>>>>> Thanks in advance,
>>>>>
>>>>> Emre
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Mary-users mailing list
>>>>> Mary-users at dfki.de
>>>>> http://www.dfki.de/mailman/cgi-bin/listinfo/mary-users
>>>>
>>>>
>>>> --
>>>> _______________________________________________________________
>>>> Marcela Charfuelan, Researcher, DFKI GmbH
>>>> Projektbuero Berlin, Alt-Moabit 91c, 10559 Berlin, Germany
>>>> Phone: +49 (0)30 39491821
>>>> URL :http://www.dfki.de/~charfuel/
>>>> _______________________________________________________________
>>>> Deutsches Forschungszentrum fuer Kuenstliche Intelligenz GmbH
>>>> Firmensitz: Trippstadter Strasse 122, D-67663 Kaiserslautern
>>>> Geschaeftsfuehrung:
>>>> Prof. Dr. Dr. h.c. mult. Wolfgang Wahlster (Vorsitzender)
>>>> Dr. Walter Olthoff
>>>> Vorsitzender des Aufsichtsrats:
>>>> Prof. Dr. h.c. Hans A. Aukes
>>>> Amtsgericht Kaiserslautern, HRB 2313
>>>> _______________________________________________________________
>>>
>>
>>
>> --
>> _______________________________________________________________
>> Marcela Charfuelan, Researcher, DFKI GmbH
>> Projektbuero Berlin, Alt-Moabit 91c, 10559 Berlin, Germany
>> Phone: +49 (0)30 39491821
>> URL :http://www.dfki.de/~charfuel/
>> _______________________________________________________________
>> Deutsches Forschungszentrum fuer Kuenstliche Intelligenz GmbH
>> Firmensitz: Trippstadter Strasse 122, D-67663 Kaiserslautern
>> Geschaeftsfuehrung:
>> Prof. Dr. Dr. h.c. mult. Wolfgang Wahlster (Vorsitzender)
>> Dr. Walter Olthoff
>> Vorsitzender des Aufsichtsrats:
>> Prof. Dr. h.c. Hans A. Aukes
>> Amtsgericht Kaiserslautern, HRB 2313
>> _______________________________________________________________
>
--
_______________________________________________________________
Marcela Charfuelan, Researcher, DFKI GmbH
Projektbuero Berlin, Alt-Moabit 91c, 10559 Berlin, Germany
Phone: +49 (0)30 39491821
URL : http://www.dfki.de/~charfuel/
_______________________________________________________________
Deutsches Forschungszentrum fuer Kuenstliche Intelligenz GmbH
Firmensitz: Trippstadter Strasse 122, D-67663 Kaiserslautern
Geschaeftsfuehrung:
Prof. Dr. Dr. h.c. mult. Wolfgang Wahlster (Vorsitzender)
Dr. Walter Olthoff
Vorsitzender des Aufsichtsrats:
Prof. Dr. h.c. Hans A. Aukes
Amtsgericht Kaiserslautern, HRB 2313
_______________________________________________________________
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.dfki.de/pipermail/mary-users/attachments/20110505/c91570b7/attachment.htm
More information about the Mary-users
mailing list