search

Kyubyong / tacotron

A TensorFlow Implementation of Tacotron: A Fully End-to-End Text-To-Speech Synthesis Model
Apache License 2.0
1.83k stars 436 forks source link

Pre-trained model sharing? #107

Closed dreamibor closed 6 years ago

dreamibor commented 6 years ago

Hi, Thank you for sharing your code, I am currently looking for a pre-trained tacotron model, it seems you have talked about this in previous issues, but I cannot find any links to pre-trained models. Did you remove that? Probably there are some legal issues? Thank you.

Kyubyong commented 6 years ago

Oh my god! Absolutely no. There's no legal issues at all when I share my pretrained model on the LJ dataset or the WEB data. I was totally forgetting it. And what is even worse is I can't find them! Let me retrain them. It should take around five days. Please check them again in five days. Sorry.

dreamibor commented 6 years ago

@Kyubyong That will be much appreciated! Thank you.

Kyubyong commented 6 years ago

I've just updated README.md with the links for pretrained models. I'm not lazy so I didn't check if they work, so please let me know if there's some problem with them.

dreamibor commented 6 years ago

@Kyubyong I tried with your newly uploaded checkpoint, and it looks good. The quality of generated voice is more than acceptable. Thank you.

dreamibor commented 6 years ago

Closed for the author has uploaded the model checkpoint.

The-Gupta commented 6 years ago

@Kyubyong ,This is not an issue, still... Could you tell a measure of accuracy for all the pretrained models with respect to the state-of-the-art TTS service available?