https://huggingface.co/hexgrad/Kokoro-82M#training-details
The model uses freely licensed audio. Nice.
But it also uses synthetic audio from closed-source models and unspecified open-source licenses. Not nice.
But is there any way I can get the dataset? If not, none of this is completely open source per OSI. The model itself is even advertised as “open-weight”, not open source.
Oh it’s just using kokoro? Lol
I don't do much TTS but is this better than the non-AI equivalent?
Maybe my ears are bad because I've listen to too much human made loud music, but it sounds more artificial than the default say command on OSX.
those samples sounded better than these ones from 2023 to me, but can we even know that apple is or isn't using AI for new versions of say?
There's quite a quality difference between the different voices.
For example in the English (UK) ones, there's 8 of them, which vary between "pretty good" and "robotic".
All about open source! Feel free to ask questions, and share news, and interesting stuff!
Community icon from opensource.org, but we are not affiliated with them.