|
From: | Marcelo Y. Matuda |
Subject: | Re: [gnuspeech-contact] GnuspeechSA 0.1.6 |
Date: | Sat, 6 May 2017 23:02:27 -0300 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.1.0 |
On 06-05-2017 21:24, David Roderic Hill wrote:
Hi Felipe, Thank you for your comment on Gnuspeech.To have Gnuspeech speak Portugese requires a significant amount of work. First Monet has to have the ability to save and retrieve database files implemented (should be pretty straightforward). Then you have to characterise the basic postures of Portugese in terms of the tube radii and test them. Then you need to create models of Portugese intonation and rhythm and test them. Has anyone worked on the rhythm and intonation of Portugese?
There are many researchers in Brazil working on portuguese speech synthesis, but many implementations are closed.
Finally you have to create a pronouncing dictionary, orthography-to-postures, for Portugese, with a method of using the dictionary to get the pronunciation of derivative words from theones actually in the dictionary, and a pre-parser to deal with standard items like dates and numbers to ensure that they are converted into a normal pronounceable form. Finally you need a set of letter-to-sound rules for Portugese in case there are words that are not found in the dictionary or obtained by the derivative procedures.Monat was designed to support such work, and it is what we used to create the databases for EnglishSimilar steps are required to create the databases for any additional language that you want Gnuspeech to speak. There are precompiled manuals Monet and TRAcT manuals on my university website give insight into what we did: http://pages.cpsc.ucalgary.caThe Latex source files for the manuals are on the Gnuspeech project FSF site. The illustrations use the Mac OSX version where screen shots are used. It took three of us rather less than a year to go from having no tools, to having created the tools, and used them to create the databases and a working system.All good wishes. david ------- David R. Hill Emeritus Professor University of Calgary Gnuspeech FSF project leader On May 6, 2017, at 15:31 13PM, Felipe Castro wrote:I like this, it' s very nice!Marcelo, have you tried GnuSpeech to speak in other languages, you seem to be brazilian, just like me, so it would nice to have this "thing" speaking in Portuguese...Felipe Castro
A long time ago I worked on a project called cgstmad. It was a simplified Gnuspeech, but it could use other programs as front-ends. I used espeak as front-end for brazilian portuguese, but the result was not very good, because I didn't tune very well the database.
I found some open source front-ends for brazilian portuguese, but I haven't tried them yet.
Marcelo
2017-05-06 19:01 GMT-03:00 Marcelo Matuda <address@hidden>: Hi,GnuspeechSA 0.1.5 has a serious (and stupid) bug in the intonation. The problem is that the intonation curve is not smooth. This bug has been fixed in GnuspeechSA 0.1.6. It can be downloaded at: https://gitlab.com/mym8/gnuspeech_sa/tags or https://github.com/mym8/gnuspeech_sa/releasesBTW there are some synthesis examples at: https://mym.eng.br/gnuspeech.htmlMarcelo_______________________________________________ gnuspeech-contact mailing list address@hidden https://lists.gnu.org/mailman/listinfo/gnuspeech-contact_______________________________________________ gnuspeech-contact mailing list address@hidden https://lists.gnu.org/mailman/listinfo/gnuspeech-contact
[Prev in Thread] | Current Thread | [Next in Thread] |