coqui-ai
/
TTS
mirror of https://github.com/coqui-ai/TTS.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'tacotron-gst' of github.com:mozilla/TTS into multispeaker

pull/10/head
Thomas Werkmeister 2019-07-01 14:00:22 +02:00
parent 765597e983 464cc29756
commit 04e452d8cb
2 changed files with 6 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
layers/common_layers.py
View File

@ -208,7 +208,7 @@ class Attention(nn.Module):
_, n = prev_alpha.max(1) _, n = prev_alpha.max(1)
val, n2 = alpha.max(1) val, n2 = alpha.max(1)
for b in range(alignment.shape[0]): for b in range(alignment.shape[0]):
alpha[b, n[b] + 2:] = 0 alpha[b, n[b] + 3:] = 0
alpha[b, :(n[b] - 1)] = 0 # ignore all previous states to prevent repetition. alpha[b, :(n[b] - 1)] = 0 # ignore all previous states to prevent repetition.
alpha[b, (n[b] - 2)] = 0.01 * val[b] # smoothing factor for the prev step alpha[b, (n[b] - 2)] = 0.01 * val[b] # smoothing factor for the prev step
# compute attention weights # compute attention weights

9
utils/audio.py
View File

@ -230,12 +230,13 @@ class AudioProcessor(object):
x = np.sign(wav) / mu * ((1 + mu) ** np.abs(wav) - 1) x = np.sign(wav) / mu * ((1 + mu) ** np.abs(wav) - 1)
return x return x
def load_wav(self, filename, encode=False): def load_wav(self, filename, sr=None):
x, sr = sf.read(filename) if sr is None:
# x, sr = librosa.load(filename, sr=self.sample_rate) x, sr = sf.read(filename)
else:
x, sr = librosa.load(filename, sr=sr)
if self.do_trim_silence: if self.do_trim_silence:
x = self.trim_silence(x) x = self.trim_silence(x)
# sr, x = io.wavfile.read(filename)
assert self.sample_rate == sr, "Expected sampling rate {} but file " \ assert self.sample_rate == sr, "Expected sampling rate {} but file " \
"{} has {}.".format(self.sample_rate, "{} has {}.".format(self.sample_rate,
filename, filename,