def predict(tst_stn):
if getattr(hps.data, "add_blank", False):
text_norm = text_to_sequence(tst_stn.strip(), ['english_cleaners'],
cmu_dict)
text_norm = commons.intersperse(text_norm, len(symbols))
else: # If not using "add_blank" option during training, adding spaces at the beginning and the end of utterance improves quality
tst_stn = " " + tst_stn.strip() + " "
text_norm = text_to_sequence(tst_stn.strip(), ['english_cleaners'],
cmu_dict)
sequence = np.array(text_norm)[None, :]
x_tst = torch.autograd.Variable(torch.from_numpy(sequence)).long()
x_tst_lengths = torch.tensor([x_tst.shape[1]])
with torch.no_grad():
noise_scale = .667
length_scale = 1.0
(y_gen_tst, *_), *_, (attn_gen, *_) = model(x_tst,
x_tst_lengths,
gen=True,
noise_scale=noise_scale,
length_scale=length_scale)
audio = waveglow.infer(y_gen_tst, sigma=.666)
audio = normalize_audio(audio[0].clamp(-1,
1).data.cpu().float().numpy())
return hps.data.sampling_rate, np.stack([audio, audio], 1)
def get_text(self, text):
text_norm = text_to_sequence(text, self.text_cleaners,
getattr(self, "cmudict", None))
if self.add_blank:
text_norm = commons.intersperse(text_norm, len(
symbols)) # add a blank token, whose id number is len(symbols)
text_norm = torch.IntTensor(text_norm)
return text_norm
def get_text(self, text):
text_norm = self.cleaner(text)
if self.add_blank:
text_norm = commons.intersperse(
text_norm, len(self.cleaner)
) # add a blank token, whose id number is len(symbols)
text_norm = torch.IntTensor(text_norm)
return text_norm
def get_text(self, text):
if self.cleaned_text:
text_norm = cleaned_text_to_sequence(text)
else:
text_norm = text_to_sequence(text, self.text_cleaners)
if self.add_blank:
text_norm = commons.intersperse(text_norm, 0)
text_norm = torch.LongTensor(text_norm)
return text_norm