Using fatchord WaveRNN Model with TTS Tacotron

Hello all,

I trained a WaveRNN model on fatchord’s repo (not the mozilla TTS one) and I was wondering if anyone had any insight in how to get it working with a TTS Tacotron model?

I am using the built in server and when trying to load the wavernn model I get the following error:

in load_wavernn
self.wavernn.load_state_dict(check[‘model’])
KeyError: ‘model’

The fatchord model generates a .pyt model while I know the pretrained models here are .pth.tar. Would that have anything to do with it?

Appreciate any help!

WaveRNN support soon to be deprecated as we introduce a new vocoder submodule under TTS. Just a small warning.