marytts
new release and build routine
hi marytts community,
it is possible to release marytts v6 beta in next time?
i see some bugfixes only there. i want build a voice from audio and text files. give it a way to build a voice full automatically? Only set input files and get the JAR voice file. Mabye someone wirte a tool or script? Sorry but the voice build routine is not easy to understand and i get some errors, see open issues. A simple tool will be great.
best regards, Paul
Hello, for v6 we massively refactored marytts. Currently, there is no backend (unit selection/spss) support but the integration is in progress. When this part will be integrated we will release a beta of marytts 6. Before, we can't as we don't have the full chain to achieve synthesis
as fast as we can, we are only 2 people to put this new version in place so I can't give a schedule. However, we are reaching the end. Like I said, we are focusing on the backend now which means everything else is in a beta mode already.
Is a feature planned, which allows to build new languages using audio and text files?
@proewer Building new languages, and building new voices (note the distinction) using audio and text files has been a feature for many years. You do not need to wait for a new release. Please have a look at the source repositories for some of the published voices to get a feeling for how this can be done automatically:
voice-cmu-slt voice-dfki-poppy voice-dfki-pavoque-neutral etc.
okay, how i can create a new voice jar file with audio and text files in this way? i don't understand how this projects work. i not see the workflow. sorry.
The READMEs for the various voice repos should be clear enough. Follow the instructions there and open new issues if you have specific questions or encounter errors. Once you understand how that works, it makes sense to prepare your own data in a similar way. The documentation on how to do this is still evolving, but in the meantime, you can have a look at these slides for further details.