I'm still chasing down an issue getting MarySpeech working and playing well on my Linux installation.

I saw a discussion on adding a language and it was talking about the mary/installed directory.  In that topic there was mention of copying an .xml file for the language to the ../mrl/mary/insatlled (I think they meant ../MRL base install dir/mary/installed) But there is a problem since that mary/installed directory does not exist on my installation.

In my case there appears to be an issue with the installation as I have the following directories (all empty):

mary\download mary\installed and mary\log in addition to the mary directory.  Under the mary directory I only have a log subdirectory.

One problem that I immediately see is that in linux the path separator is '/' and NOT '\' so that is suspect.

Under libraries/jar I have a number of marytts-*.jar files including for various languages.

If you have MarySpeech installed under Linux do you have the same invalid empty directories or is this something about my installation here?

Thanks,

Burt

 

burtbick

7 years ago

Some more information:

I went back and ran MRL 1.0.2000 using the reinstalled Java 8.  I noticed that the maryTTS installation there had the same bogus directories, but it DID have a mary/installed directory that contained the .xml files for the installed parts of maryTTS.  That directory is missing from the newer installation I did under a separate directory path.

I am able to execute the simple test script on that 1.0.2000 MRL installation and other than incredibly bad voice quality it works with no exceptions.

So has anyone else seen installation issues with maryTTS that would explain the issues that I'm seeing.  Just trying to understand where it broke so I can try to avoid similar issues in the future.

Thanks,

Burt

AutonomicPerfe…

7 years ago

I seem to be having similar issues. Running a script to change the voice in mrl version 1.0.1758 works fine, but replacing the myrobotlab.jar file with the latest version in the same directory produces the error that the voice cannot be found. So, something in the code must have been changed between then and now. I would send a no-worky, but I am away from my workstation right now.

As to the directory structures, I as well see the same thing, with the mary\installed etc. directories only being created with the older version. With the newest version, no directories related to maryspeech are created at all, which could be why the voice files aren't being recognized. I am using Ubuntu 16.04, and on Windows the directory structure is created correctly with the old version, implying that the problem is someone used a hard-coded "\" instead of File.separator. Still, nothing is created under the newest version, though.

Hope it helps.

Gradually getting it working.  I went back to 1.0.2000 (latest non-stable I believe) and on one of my installs I wound up with a mary/installed directory there that was properly populated. 

With that installation I'm able to run a simple script to say Hello World.  I need to figure out the best way to load some new voices to experiment with as the default English voice is pretty lame, but it works for testing.

Also I found in the MRL Browser based interface for Python the Examples list has a MarySpeech script example that would have been helpful had I found it earlier before Ken W provided me with a simple script, so there and pyrobotlab under MyRobotLab on GitHub has a bunch of different Python scripts for various functions which will be my goto place at least until I hit a speedbump with one of them and the newer versions of MRL.  Note this doesn't appear in the GUI application of MRL.  That is the examples option brings up a github page but I don't see a MarySpeech example there.

It sounds like you had an issue with the install failing to fully install MarySpeech on Linux, and I have had that issue happening.  I'm not sure yet what the secret sauce is to get a proper installation, but obviously one time I DID get MarySpeech to install well enough to run.

Do you have a "mary" directory in your installation, or no "mary anything" directories? 

Did you try a fresh installation of the latest MRL jar in a fresh directory?  That was what I wound up doing.  That might be worth a try.

 

Burt

I'm not using MarySpeech much as I found the voice very ugly. Too bad AcapellaSpeech and NaturalReaderSpeech stop working. But there is a few recent speech service that seem to give better result than Mary

About Mary, the version use in MRL was upgrade to 5.2(?) between 1758 and the latest build. That's probably why your voice was working in 1758 but not in the latest build. There was a blog post recently on this site about how to make it work. Look for it and youwill probably be able to fix your problem

 

Been busy cleaning up the Inmoov head and a test stand that I built so I can experiment more with tracking, as well as voice commands and text to speech.

I was able to get it to work with the latest build as mentioned above.  Something had happened with the last Linux Mint 17.1 update that seemed to have broken the Oracle Java installation.  I went back to OpenJDK 8 and was able to get the latest build to work with MarySpeech.

Thanks,

Burt