is this right link of the pretrained model 1000k?

[Yeongtae]

@r9y9 Using above link, I could get bellow pretrained model.

• It seems to be not 100K.

I have run synthesize.py using the pretrained model at the above image. But I could get a noisy wav file.

0817.zip

The wavs with prefix 'wav-' are generated by using https://github.com/Rayhane-mamah/Tacotron-2 with your pretrained model A wav with prefix '20180510_' is generated by using above pretrained model.

Please check above link.

[r9y9]

Have you checked https://github.com/r9y9/wavenet_vocoder#online-tts-demo? Let me know if it doesn't work.

[r9y9]

It seems to be not 100K.

This is because I trained the model multiple times from step 0. Approx 300k * 4 > 1000k..

[Rayhane-mamah]

Hello, pardon me intruding here :)

@Yeongtae you don't seem to be rescaling T2 output mels to [0, 1] @r9y9's demo shows how to do that correctly. I'm out of here :)

[r9y9]

I'm closing this. Feel free to reopen if the problem persists.

[Yeongtae]

@r9y9 Thanks