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

Add new gl implementations

This commit is contained in:
Eren Golge 2018-04-17 09:57:57 -07:00
parent ee32fbc011
commit e2d974e8be
1 changed files with 21 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
utils/audio.py
View File

@ -74,16 +74,29 @@ class AudioProcessor(object):
# Reconstruct phase # Reconstruct phase
return self.apply_inv_preemphasis(self._griffin_lim(S ** self.power)) return self.apply_inv_preemphasis(self._griffin_lim(S ** self.power))
# def _griffin_lim(self, S):
# '''librosa implementation of Griffin-Lim
# Based on https://github.com/librosa/librosa/issues/434
# '''
# angles = np.exp(2j * np.pi * np.random.rand(*S.shape))
# S_complex = np.abs(S).astype(np.complex)
# y = self._istft(S_complex * angles)
# for i in range(self.griffin_lim_iters):
# angles = np.exp(1j * np.angle(self._stft(y)))
# y = self._istft(S_complex * angles)
# return y
def _griffin_lim(self, S): def _griffin_lim(self, S):
'''librosa implementation of Griffin-Lim '''Applies Griffin-Lim's raw.
Based on https://github.com/librosa/librosa/issues/434
''' '''
angles = np.exp(2j * np.pi * np.random.rand(*S.shape)) S_best = copy.deepcopy(S)
S_complex = np.abs(S).astype(np.complex)
y = self._istft(S_complex * angles)
for i in range(self.griffin_lim_iters): for i in range(self.griffin_lim_iters):
angles = np.exp(1j * np.angle(self._stft(y))) S_t = self._istft(S_best)
y = self._istft(S_complex * angles) est = self._stft(S_t)
phase = est / np.maximum(1e-8, np.abs(est))
S_best = S * phase
S_t = self._istft(S_best)
y = np.real(S_t)
return y return y
def melspectrogram(self, y): def melspectrogram(self, y):
@ -97,7 +110,7 @@ class AudioProcessor(object):
def _istft(self, y): def _istft(self, y):
_, hop_length, win_length = self._stft_parameters() _, hop_length, win_length = self._stft_parameters()
return librosa.istft(y, hop_length=hop_length, win_length=win_length) return librosa.istft(y, hop_length=hop_length, win_length=win_length, window='hann')
def find_endpoint(self, wav, threshold_db=-40, min_silence_sec=0.8): def find_endpoint(self, wav, threshold_db=-40, min_silence_sec=0.8):
window_length = int(self.sample_rate * min_silence_sec) window_length = int(self.sample_rate * min_silence_sec)