acetylsv / gst-tacotron Goto Github PK

If possible then please share pretrained models of respective datasets.

How do we ensure that the padding of the reference mel spectogram is taken into account when the reference encoder is applied on a batch of mels?

Hi,

For condition_on_audio= False case, How to compute style_emb. What should be the GST tokens ?

It seems that the link to the pre-trained models is down. I was just wondering if you are able to put them up again. Thanks in advance of any help.

Thanks for sharing pre-trained models. Could you please let know how to use it.. If I rand it is asking me about input data set..

Hi,

I found this project is fantastic, thanks your contribution.

but I face some problem on trying to inference by GST tokens.

Already trained a model with LJSpeech, set condition_on_audio = False, but infer.py will output error as below:

ValueError: operands could not be broadcast together with shapes (320,) (256,)

I knew shapes (320,) and (256,) not aligned，but how to fix by modify your code to solve it?

if you can give some suggestions will be grateful. Thanks.

Hello! I have some questions about the BZ datasets. Do you have data preprocessing operation on the BZ dataset before training the model, such as breaking long sentences into small segments? Some sentences in BZ datasets are much longer than sentences in LJ.

I use pre-trained models and different reference audio, but the resulting audio talks barely change.
What could be the reason for this?