hexgrad

/

Kokoro-82M

like 3.15k

Yes, would be awesome to have also German voices!

I second that.

Here are also 23+ hours of german voice with permissive license https://www.thorsten-voice.de/datasets/

I third that

+1 for german support, would be amazing to add it to the list of voices

Hi,
is there a chance to get a german voice for kokoro?

Yes +1 for German. But for general purposes I think we need better samples. The ones posted here as an example sound more like for book audio readings.

german is the only voice preventing me to use it, would be great if you can add it!

I can only add eligible audio that I find, or that others give to me. So far, I have received no German audio in response to https://hf.co/posts/hexgrad/846477530846098

Edit: there are quality standards for audio that enters training, and things like Common Voice or Thorsten Voice do not meet that bar

Are you aware of the following datasets?

•	M-AILABS German and CSS10 – German for additional single or multi–speaker data. https://github.com/imdatceleste/m-ailabs-dataset
•	HUI-Audio-Corpus-German for a large, multi–speaker resource with excellent audio quality: https://github.com/iisys-hof/HUI-Audio-Corpus-German
•	LibriVoxDeEn for audiobook–based, read speech data: https://www.cl.uni-heidelberg.de/statnlpgroup/librivoxdeen/

@hexgrad
here is dataset created from the OpenAI TTS API: https://huggingface.co./datasets/laion/laions_got_talent
there is a lot of german audio in this dataset.

Would that data qualify for your training?