[mary-users] Compressed Output?

Alex Shaykevich ashayk1 at yahoo.com
Mon Feb 1 16:04:07 CET 2010


No worries.  Though, seriously, Speex is BY FAR the best option, especially if you've got control over the decoding on the back end.  Speex is optimized for the human voice frequency range and sounds fantastic.  Plus, as I said, there is a pure Java port.  I have used it with Mary output and the results are excellent.

I recently tried EVERYTHING, including re-implementing the Quicktime IMA4 C code you're talking about as a JMF codec, and I simply couldn't get the last bit of fuzz out.  Like I said, the ffmpeg crew haven't managed it either and I have an outstanding bug.  If you managed to get clean IMA4 compression I will be VERY interested in testing it because I could use one on Windows.  On OSX, afconvert will create AAC or IMA4 readily using Apple's codecs which are the only ones I've ever heard that create pristine IMA4.  As you probably know, there is actually some variability in IMA4 compression, and I think Apple does something unusual either in their codec or, I suspect, in the AIFC file writer.


Good luck,

Alex


________________________________
From: Clifton Craig <clifton.craig at gmail.com>
To: Alex Shaykevich <ashayk1 at yahoo.com>
Cc: mary-users at dfki.de
Sent: Tue, February 2, 2010 1:49:10 AM
Subject: Re: [mary-users] Compressed Output?

Thanx Alex,

I realized how silly I sounded suggesting JOgg and JavaZoom shortly after sending the email. I'm looking into IMA4 because its a format I can handle easily on the client end. I'm constrained by device limitations and legal technicalities for the work I'm doing so MP3 is actually out of the question. I found the java-vorbis trunk but due to the questionable quality I may defer to a native library. Because I've never heard of Speex I'm apprehensive. I'd probably need to roll my own decoder on the client. So it looks like IMA4 may be the best bet (assuming speech compression doesn't run into the noise you'd get with music under the same encoding) with OGG as a second option.


Clifton Craig
Software Engineer
http://codeforfun.wordpress.com
clifton.craig at gmail.com 

On Feb 1, 2010, at 8:56 AM, Alex Shaykevich wrote:

JLayer is an MP3 decoder, but not encoder. Likewise JOgg, though somewhere there is an experimental Java Ogg encoder.  The best codec for Mary output would have to be Speex, and JSpeex is a pure Java encoder.  The downside with Speex, however, is that WMP and Quicktime will not play Speex files without plugins, but it depends what your goals from the compression ultimately are.  Speex will produce the smallest files with the best audio quality for speech since it's a speech codec from the getgo.
>
>BTW, JMF has an IMA4 codec in it and you can create AIFC with IMA4 compression.  For music, I found it's audio quality subpar, though ffmpeg's IMA4 encoding is equally noisy.  I've filed a bug with the ffmpeg crew but they're in no hurry.
>
>Lastly, if you're on Windows, Tritonus.org provides an mp3 encoder through JNI using LAME.
>
>--Alex
>
>
>
>
________________________________
From: Clifton Craig <clifton.craig at gmail.com>
>To: Marc Schroeder <schroed at dfki.de>
>Cc: mary-users at dfki.de
>Sent: Tue, February 2, 2010 12:37:25 AM
>Subject: Re: [mary-users] Compressed Output?
>
>Marc,
>
>I'd volunteer if I had some help figuring out where to put such a  
>change. First thing's first. We have to find a solution. Would JLayer  
>from JavaZoom be a possibility? I'm also looking at JOgg. I've also  
>found the source code to IMA4 in C. I was thinking of attempting a  
>Java port. I'm going to try porting/using the IMA4 library first and  
>see how far I get.
>
>Clifton Craig
>Software Engineer
>http://codeforfun.wordpress.com
>clifton.craig at gmail.com
>
>On Feb 1, 2010, at 4:08 AM, Marc Schroeder wrote:
>
>> Clifton,
>>
>> MARY 4 currently doesn't support any compressed formats. The core  
>> reason
>> is that we worked hard to make MARY 4 Java-only, and there seem to  
>> be no
>> Java-only encoders for mp3 or ogg available, or they are proofs of
>> concept only.
>>
>> I have been thinking that an alternative would be to allow for an
>> external compression executable to be called after the audio is
>> synthesised and before it is sent to the client. Since there are no
>> resources to add this kind of support at our end, this job can only be
>> done by a volunteer out there, willing to agree with me on the right
>> place to put this. Anybody interested?
>>
>> Regards,
>> Marc
>>
>> On 29.01.10 15:18, Clifton Craig wrote:
>>> Where can I find examples on how to configure Mary for compressed
>>> output? Out
>>> of the box Mary seems to only support AU, WAVE, and AIFF formats.  
>>> There are
>>> hints of MP3 and OGG support in the Javadocs for MaryServer
>>> (http://mary.dfki.de/javadoc/4.0%20beta/marytts/server/http/MaryHttpServer.html 
>>> )
>>>
>>> but I don't see anything that illustrates how to enable it.
>>>
>>> Clifton Craig
>>> Software Engineer
>>> http://codeforfun.wordpress.com
>>> clifton.craig at gmail.com <mailto:clifton.craig at gmail.com>
>>>
>>>
>>>
>>> _______________________________________________
>>> Mary-users mailing list
>>> Mary-users at dfki.de
>>> http://www.dfki.de/mailman/cgi-bin/listinfo/mary-users
>>
>> -- 
>> Dr. Marc Schröder, Senior Researcher at DFKI GmbH
>> Coordinator EU FP7 Project SEMAINE http://www.semaine-project.eu
>> Project leader for DFKI in SSPNet http://sspnet.eu
>> Project leader PAVOQUE http://mary.dfki.de/pavoque
>> Associate Editor IEEE Trans. Affective Computing http://computer.org/tac
>> Editor W3C EmotionML Working Draft http://www.w3.org/TR/emotionml/
>> Portal Editor http://emotion-research.net
>> Team Leader DFKI TTS Group http://mary.dfki.de
>>
>> Homepage: http://www.dfki.de/%7Eschroed
>> Email: schroed at dfki.de
>> Phone: +49-681-302-5303
>> Postal address: DFKI GmbH, Campus D3_2, Stuhlsatzenhausweg 3, D-66123
>> Saarbrücken, Germany
>> --
>> Official DFKI coordinates:
>> Deutsches Forschungszentrum fuer Kuenstliche Intelligenz GmbH
>> Trippstadter Strasse 122, D-67663 Kaiserslautern, Germany
>> 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
>> _______________________________________________
>> Mary-users mailing list
>> Mary-users at dfki.de
>> http://www.dfki.de/mailman/cgi-bin/listinfo/mary-users
>
>_______________________________________________
>Mary-users mailing list
>Mary-users at dfki.de
>http://www.dfki.de/mailman/cgi-bin/listinfo/mary-users
>
>_______________________________________________
>Mary-users mailing list
>Mary-users at dfki.de
>http://www.dfki.de/mailman/cgi-bin/listinfo/mary-users
>



      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.dfki.de/pipermail/mary-users/attachments/20100201/70e18ad2/attachment.htm 


More information about the Mary-users mailing list