Add languages

[nozem79]

I do not know how to properly ask a question in Github, so I hope this is the right way.
Since I presume the voices are based on WAV's how difficult is it to ad a language like dutch & what is te proper way to start do do that?

Kind regardsm

Daniel

[jpwhiting]

Hello Daniel,

Actually, the languages are based on some data files that are in a format that is known, but generated by some other tools that aren't included in this project. There are a dozen or so files, mostly lists of numbers for signal generation, phoneme pronunciation, etc. etc. Included in upstream git repository are windows only binary tools that convert those into data files that pico library can read and use. In this other fork I created here: https://github.com/jpwhiting/svox-pico I have tried to reverse engineer some of those tools and rewrite them in python so they will work on linux/mac also, but haven't got them all yet. Once those are all converted and rewritten we will be able to generate the existing data files for the existing languages. Understanding the source data files would be required to add more languages. I hope we get to that at some point, but I think it's going to be far off.

The other problem is that the company that created this (SVOX) has been bought by Google last decade so there's not really a known person or place to reach out to for guidance in figuring this out as far as I can tell.

[nozem79]

Thanks fot the swift replay, I hope it's possible to use Pico in my language of choice.

[ihuguet]

@jpwhiting, although having those tools as free software would be great, in the meanwhile those binaries can be used to generate the languages, is that correct?

What's upstream git repo? I just got this sources from debian's package.

[jpwhiting]

Yes, those windows tools work fine for now. I think they are preventing debian from putting pico in main though, they are in non-free because those exe files are an unknown I guess https://packages.debian.org/bullseye/libttspico0

Upstream repo for svox-pico is here: https://android.googlesource.com/platform/external/svox/ the pdfs documenting the data files are on donut-release branch iirc.

[nozem79]

Somehow I cannot find the exe's , then again, I'm half blind ;-)

[jpwhiting]

Daniel, They are here: https://android.googlesource.com/platform/external/svox/+/refs/heads/donut-release But they only convert the textfiles to .pkb then there's a perl script to combine those into .bin files 2 per language, one for the textanalysis bits and the other for the sound generation bits. The trick to add a new language is to come up with the text files used as the starting point for generating each of those. That commit also has some useful pdfs about how it all works, but there's not much about where the data files came from or how to add new ones unfortunately. BR, Jeremy

…

On Wed, Jan 12, 2022 at 4:28 AM Daniel ***@***.***> wrote: Somehow I cannot find the exe's , then again, I'm half blind ;-) — Reply to this email directly, view it on GitHub <#7 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAHDYPELSOGAF5ISNAUSCY3UVVQULANCNFSM5LO6CEZA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[versae]

I was looking at the outputs of pico and they look great, but was wondering how to add other languages and then found this issue. Any luck at all adding a new language?