Is there any plan for supporting other language?
Hi, thanks for great work.
I just test it my reference audio using your space, it was awesome. In my experience, this is the best model for AI japanese waifu.
Despite of great quality of your tts model, For AI waifu that handle other things like JARVIS, it would be great support other language like english.
Do you have a any plan for supporting other language?
Thanks for sharing the model :)
Thank you so much!
Architecturally speaking, supporting English is no problem at all. All it takes is high-quality datasets and compute resources.
However, since I'm Japanese, I'm not very familiar with what makes a good English dataset. Additionally, adding data for other languages means expanding the overall dataset size, which inevitably drives up the training cost.
Because of these reasons, I don't have any plans to support other languages at the moment. But thank you again for your kind words and for testing it out!
what kind of resources do you need ?? and English version will be fantastic, echoTTS is too big
is that kind of quality in 500M model ?
or maybe release the training pipeline...
The training code is actually already available here!
If you'd like to try training it yourself, please feel free to check it out:
https://github.com/Aratako/Irodori-TTS
what are the resources used for training Irodori ?