Alexey322

Rayhane-mamah Tacotron-2 + r9y9 wavenet_vocoder compatibility

1

Hey. I trained the Tacotron 2 synthesizer from Rayhane-mamah, the synthesized spectrograms sound good if you use the griffin limm algorithm. Unfortunately, the vocoder in his repository learns with an...

Hey. I trained the Tacotron 2 synthesizer from Rayhane-mamah, the synthesized spectrograms sound good if you use the griffin limm algorithm. Unfortunately, the vocoder in his repository learns with an...

Pad audio fragment

2

Why do we need pad audio fragment while receiving its mel spec? `y = torch.nn.functional.pad(y.unsqueeze(1), (int((n_fft-hop_size)/2), int((n_fft-hop_size)/2)), mode='reflect')`

Inconsistency between the model parameters in the paper and the implementation on the github

2

@jik876 Hi. I would like to know why you are not using the same parameters(for V1 configuration) as indicated in the paper? Your code has set the following parameters: "resblock_kernel_sizes":...

What are the correct model changes for sample rate 44100?

1

Hi, @jik876. Can you give some advice on how to change the model correctly for the 44100 sample rate? I don't mean hyperparameters in config. For example, how did you...

ExponentialLR and fine tune

3

Hi. I started training the model from scratch and found that the optimizer uses a dynamic learning step. If I train the model with 2.5 million steps, then according to...

Training loss becomes nan when the number of speakers changes

3

Hi. I trained the flowtron on two speakers, for a total of 50 hours, 25 for each. After that, I wanted to train the model for 10 speakers for 20-30...

Hello. Why the attention doesn't use speaker embeddings to find the alignment between text and mel spectrograms? This can vary greatly between speakers speaking in different styles.