no libre software realistic text to speech voice?

2 respostas [Última entrada]
tonlee
Desconectado
Joined: 09/08/2014

On windows 7 I used voice anna for text to speech. I also found other languages.
All were very realistic.

Those free software voices I find are not good at all.
Why are there no realistic free software text to speech voices? Is the technology to complicated?
I have found http://www.cepstral.com/en/personal/download/?os=linux&refpage=linux voices. I do not know the license. I have not been able to install them. The demo voices are close to acceptable.

https://www.youtube.com/watch?v=UzaqNF6NlBA Do you want to use it with that voice? I would want to avoid that voice.

megurineturilli
Desconectado
Joined: 01/10/2012

No, there is no free software yet that can do this. Cepstral. MBROLA and many others are non-free. I am currently working on a free software that does the job. It is based on WORLD (https://github.com/mmorise/World) which is already free. The hardest part is creating the database.

tonlee
Desconectado
Joined: 09/08/2014

>database

Is it samples of voice?
This is another example of libre software not being able to do a campaign. 5 major languages are not available in an adequate quality. But libre software people are not able to mobilize the money or a campaign or skills to make it.
If you are making such an important piece of software by yourself, it tells a lot.

Now I may get the comments saying make it yourself. No, I cannot. I do not have the skills or money for it. I do have 10eu for it, if there would be a crowd funding being run by an libre software entity known to be valid. Then I would not have to validate the crowd funding myself. I would only have to decide if a given crowd funding would be something I would support. Unless fsf does not have the resources to frontrun crowd fundings, I find it incomprehensible that the fsf do not. Or the banal answer is, free software people are not going to support crowd fundings in significant numbers.