[mary-users] Where can I find different voices? Feeling very helpless

[Ingmar Steiner]

Dear Mathias,

sorry to read about your frustrating experience so far. Let me shed some 
light:

1) The main website has always been, and remains, at http://mary.dfki.de/.

2) All development is at https://github.com/marytts/marytts.

3) You mention http://mary.opendfki.de/ but if you surf there, you'll 
read that that's the old development site, frozen as of June 2012, when 
development was migrated to GitHub.

Having said that, we recently redesigned the MaryTTS installer to be as 
hassle-free as possible in a very small package:

https://github.com/marytts/marytts/releases/download/v5.1.2/marytts-installer-5.1.2.zip

Unzip, and run the marytts wrapper script. Then surf to 
http://localhost:59125.

To answer your question, voices are hosted on mary.dfki.de and a Dropbox 
mirror. The new installer leverages both of these as remote repositories 
and retrieves the voice artifacts from there.

MaryTTS is primarily developed as a research tool, but we do put a lot 
of thought into making it as user- and developer-friendly as possible, 
and implement improvements when we find the time. But please bear in 
mind that it's constantly evolving.

Feel free to report any issues at 
https://github.com/marytts/marytts/issues or post general discussion 
here on the user list.

Best wishes,

-Ingmar

On 3/4/15 10:51 PM, me at mathiassteiger.biz wrote:
>
> Excuse me but, after over an hour of searching on google and scruffing through several sites, documentation, etc.  I am really really frustrated and alienated how difficult it is to publicly access new voices for mary-tts.
>
> I found several voices now (bits3-hsmm, bits1-hsmm, istc-lucia-hsmm), but they were hidden in impossible locations: Some entirely unrelated personal blogpost documentation that contained an unlisted link by accident off-site, then in a github ticket I found a link to some repository download folder on https://mary.opendfki.de which is password-protected ...
>
>
>
> I have experience with festival already, building new voices. It looks like this: You spent 11 hours solid work on reading docs and googling how to get all the 3 dozen configuration parameters straight on your system that happen to not be wrong or not set at all by default. The example commands do not work, you have to fix advanced bash scripts that misfit your system or the packages you work with, etc. You write patches for C code. Everything you read assumes that you are very knowledgable with all the 10 or so different technologies the software is build on. Then you compile the voice for 7 days straight at maximum CPU on some very recent processor (like i5-2500k), producing over a hundred Gigabyte of data, only to realize that the compilation does not finish...
>
> You give up, because it is too insane. Later you find all the voices readily available on the festivox demo site and ask them to just make them available for download. They deny the request, for strange reasons. Then after 2 years some 12MB package on some distribution, that only works exactly with the installation on this distribution, suddenly becomes available and has all the voices ... so you run it in a virtual machine.
>
>
> Why is this so? I understand that for marytts I would have to equally go through this compilation process. Why do the people that wrote it and tested it not just upload it somewhere? Strange copyright issues? I have only ever experienced such difficulties with aggressively DCMA-ed fringe-purpose software. I am a sysadmin and very knowledgable with linux, programming, etc. But this just blows my mind.
> _______________________________________________
> Mary-users mailing list
> Mary-users at dfki.de
> http://www.dfki.de/mailman/cgi-bin/listinfo/mary-users
>