-
#### Description
I believe this is how `pitch_shift` works, but I don't know much about it. As titled, in the below are three log Mel-spectrogram; from left to right: original, up-shifted by 2 steps,…
-
This is an umbrella issue to track progress for my planned TODOs. Comments and requests are welcome.
### Goal
- [x] achieve higher speech quality than conventional vocoder (WORLD, griffin-lim, e…
-
Once all dependencies have been fully installed, the system does build on Fedora 28 (it will not build on Fedora 29 due to Meson version issues), but none of the effects do anything.
-
```python
def parse_audio( y):
n_fft = int(16000 * 0.02)
win_length = n_fft
hop_length = int(16000 * 0.01)
# print(hop_length)
# STFT
D = librosa.stft(y, n_fft=n_fft…
-
I was digging into an issue where pressing `RET` on some entries gave an error, and got far enough to realize the issue is that `org-ref-key-in-file-p` wasn't finding the key properly if the entry typ…
ssfrr updated
5 years ago
-
Some electrical sounds exist in the generated audio, which greatly affects the sense of hearing.
In the original text, there is some phase loss added to the loss. Do you have any thoughts of it?
-
looking to your proposal, @ksanjeevan, I think we need to write-up some arguments for the changes. For example, I really like your `spectrogram` function and the `STFT` wrapper. Basically torch audio …
-
Thank you for this awesome work.
When I am trying to resume my training phase 2, an error occurs that
`RuntimeError: Could not read model from data/models/rnn.network`
I check the code and fou…
-
Hi，
Thanks for sharing this project ！
I have tried locally to train but the due to the slow GPU，It runs much slow. Is there any pretrained model or sample wav could be shared for the Tacotron 2 pa…
-
Hello,
So after successfully having run merlin on CPU, I would like to switch to my GPU, hereby allowing for faster training.
The problem is that _merlin_ is able to detect the GPU (Running on G…