def _prepare_targets(self, targets, maxlen):
#[batch_size, time_steps]
if is_mulaw_quantize(self._hparams.input_type):
y_batch = np.stack([_pad_targets(x, maxlen)
for x in targets]).astype(np.int32)
else:
y_batch = np.stack([_pad_targets(x, maxlen)
for x in targets]).astype(np.float32)
assert len(y_batch.shape) == 2
#Add extra axis (make 3 dimension)
y_batch = np.expand_dims(y_batch, axis=-1)
return y_batch
def _prepare_inputs(self, inputs, maxlen):
if is_mulaw_quantize(self._hparams.input_type):
#[batch_size, time_steps, quantize_channels]
x_batch = np.stack([
_pad_inputs(
np_utils.to_categorical(
x, num_classes=self._hparams.quantize_channels),
maxlen) for x in inputs
]).astype(np.float32)
else:
#[batch_size, time_steps, 1]
x_batch = np.stack([
_pad_inputs(x.reshape(-1, 1), maxlen) for x in inputs
]).astype(np.float32)
assert len(x_batch.shape) == 3
#Convert to channels first [batch_size, quantize_channels (or 1), time_steps]
x_batch = np.transpose(x_batch, (0, 2, 1))
return x_batch
def _process_utterance(mel_dir, linear_dir, wav_dir, index, wav_path, text,
hparams):
"""
Preprocesses a single utterance wav/text pair
convert audio data to:
- audio time serie data form (numpy array)
- mel + linear spectrogram matrix (numpy matrix)
this writes the mel scale spectogram to disk and return a tuple to write
to the train.txt file
Args:
- mel_dir: the directory to write the mel spectograms into
- linear_dir: the directory to write the linear spectrograms into
- wav_dir: the directory to write the preprocessed wav into
- index: the numeric index to use in the spectogram filename
- wav_path: path to the audio file containing the speech input
- text: text spoken in the input audio file
- hparams: hyper parameters
Returns:
- A tuple: (audio_filename, mel_filename, linear_filename, time_steps, mel_frames, linear_frames, text)
"""
try:
# Load the audio as numpy array
wav = load_wav(wav_path, sr=hparams.sample_rate)
except FileNotFoundError: # catch missing wav exception
print(
'file {} present in csv metadata is not present in wav folder. skipping!'
.format(wav_path))
return None
# rescale wav
if hparams.rescale:
wav = wav / np.abs(wav).max() * hparams.rescaling_max
# check for M-AILABS extra silence specific
if hparams.trim_silence:
wav = trim_silence(wav, hparams)
# check for Mu-law quantize
if is_mulaw_quantize(hparams.input_type):
# [0, quantize_channels)
out = mulaw_quantize(wav, hparams.quantize_channels)
# Trim silences
start, end = start_and_end_indices(out, hparams.silence_threshold)
wav = wav[start:end]
out = out[start:end]
constant_values = mulaw_quantize(0, hparams.quantize_channels)
out_dtype = np.int16
#check for mu_larw
elif is_mulaw(hparams.input_type):
# [-1, 1]
out = mulaw(wav, hparams.quantize_channels)
constant_values = mulaw(0., hparams.quantize_channels)
out_dtype = np.float32
else:
# [-1, 1]
out = wav
constant_values = 0.
out_dtype = np.float32
# Compute the mel scale spectrogram from the wav
mel_spectrogram = audio_series_to_mel(hparams, wav).astype(np.float32)
mel_frames = mel_spectrogram.shape[1]
print('Debug: audiopreprocessing.py 214 mel_frames={}'.format(mel_frames))
print('Debug: audiopreprocessing.py 214 mel_spectrogram.shape={}'.format(
mel_spectrogram.shape))
print('Debug: audiopreprocessing.py 214 max_mel_frames={}'.format(
hparams.max_mel_frames))
if mel_frames > hparams.max_mel_frames and hparams.clip_mels_length:
return None
# Compute the linear scale spectrogram from the wav
linear_spectrogram = audio_series_to_linear(hparams,
wav).astype(np.float32)
linear_frames = linear_spectrogram.shape[1]
# sanity check
assert linear_frames == mel_frames
# Ensure time resolution adjustement between audio and mel-spectrogram
fft_size = hparams.n_fft if hparams.win_size is None else hparams.win_size
l, r = pad_lr(wav, fft_size, get_hop_size(hparams))
# Zero pad for quantized signal
out = np.pad(out, (l, r), mode='constant', constant_values=constant_values)
assert len(out) >= mel_frames * get_hop_size(hparams)
# time resolution adjustement
# ensure length of raw audio is multiple of hop size so that we can use
# transposed convolution to upsample
out = out[:mel_frames * get_hop_size(hparams)]
assert len(out) % get_hop_size(hparams) == 0
time_steps = len(out)
# Write the spectrogram and audio to disk
audio_filename = 'speech-audio-{:05d}.npy'.format(index)
mel_filename = 'speech-mel-{:05d}.npy'.format(index)
linear_filename = 'speech-linear-{:05d}.npy'.format(index)
np.save(os.path.join(wav_dir, audio_filename),
out.astype(out_dtype),
allow_pickle=False)
np.save(os.path.join(mel_dir, mel_filename),
mel_spectrogram.T,
allow_pickle=False)
np.save(os.path.join(linear_dir, linear_filename),
linear_spectrogram.T,
allow_pickle=False)
# Return a tuple describing this training example
return (audio_filename, mel_filename, linear_filename, time_steps,
mel_frames, text)
def _process_utterance(mel_dir, wav_dir, index, wav_path, hparams):
"""
Preprocesses a single utterance wav/text pair
this writes the mel scale spectogram to disk and return a tuple to write
to the train.txt file
Args:
- mel_dir: the directory to write the mel spectograms into
- linear_dir: the directory to write the linear spectrograms into
- wav_dir: the directory to write the preprocessed wav into
- index: the numeric index to use in the spectrogram filename
- wav_path: path to the audio file containing the speech input
- text: text spoken in the input audio file
- hparams: hyper parameters
Returns:
- A tuple: (audio_filename, mel_filename, linear_filename, time_steps, mel_frames, linear_frames, text)
"""
try:
# Load the audio as numpy array
wav = audio.load_wav(wav_path, sr=hparams.sample_rate)
except FileNotFoundError: #catch missing wav exception
print(
'file {} present in csv metadata is not present in wav folder. skipping!'
.format(wav_path))
return None
#M-AILABS extra silence specific
if hparams.trim_silence:
wav = audio.trim_silence(wav, hparams)
#Pre-emphasize
preem_wav = audio.preemphasis(wav, hparams.preemphasis,
hparams.preemphasize)
#rescale wav
if hparams.rescale:
wav = wav / np.abs(wav).max() * hparams.rescaling_max
preem_wav = preem_wav / np.abs(preem_wav).max() * hparams.rescaling_max
#Assert all audio is in [-1, 1]
if (wav > 1.).any() or (wav < -1.).any():
raise RuntimeError('wav has invalid value: {}'.format(wav_path))
if (preem_wav > 1.).any() or (preem_wav < -1.).any():
raise RuntimeError('wav has invalid value: {}'.format(wav_path))
#Mu-law quantize
if is_mulaw_quantize(hparams.input_type):
#[0, quantize_channels)
out = mulaw_quantize(wav, hparams.quantize_channels)
#Trim silences
start, end = audio.start_and_end_indices(out,
hparams.silence_threshold)
wav = wav[start:end]
preem_wav = preem_wav[start:end]
out = out[start:end]
constant_values = mulaw_quantize(0, hparams.quantize_channels)
out_dtype = np.int16
elif is_mulaw(hparams.input_type):
#[-1, 1]
out = mulaw(wav, hparams.quantize_channels)
constant_values = mulaw(0., hparams.quantize_channels)
out_dtype = np.float32
else:
#[-1, 1]
out = wav
constant_values = 0.
out_dtype = np.float32
# Compute the mel scale spectrogram from the wav
mel_spectrogram = audio.audio_series_to_mel(hparams,
preem_wav).astype(np.float32)
mel_frames = mel_spectrogram.shape[1]
if mel_frames > hparams.max_mel_frames and hparams.clip_mels_length:
return None
if hparams.use_lws:
#Ensure time resolution adjustement between audio and mel-spectrogram
fft_size = hparams.n_fft if hparams.win_size is None else hparams.win_size
l, r = audio.pad_lr(wav, fft_size, audio.get_hop_size(hparams))
#Zero pad audio signal
out = np.pad(out, (l, r),
mode='constant',
constant_values=constant_values)
else:
#Ensure time resolution adjustement between audio and mel-spectrogram
l_pad, r_pad = audio.librosa_pad_lr(wav, hparams.n_fft,
audio.get_hop_size(hparams))
#Reflect pad audio signal (Just like it's done in Librosa to avoid frame inconsistency)
out = np.pad(out, (l_pad, r_pad),
mode='constant',
constant_values=constant_values)
assert len(out) >= mel_frames * audio.get_hop_size(hparams)
#time resolution adjustement
#ensure length of raw audio is multiple of hop size so that we can use
#transposed convolution to upsample
out = out[:mel_frames * audio.get_hop_size(hparams)]
assert len(out) % audio.get_hop_size(hparams) == 0
time_steps = len(out)
# Write the spectrogram and audio to disk
audio_filename = os.path.join(wav_dir, 'audio-{}.npy'.format(index))
mel_filename = os.path.join(mel_dir, 'mel-{}.npy'.format(index))
np.save(audio_filename, out.astype(out_dtype), allow_pickle=False)
np.save(mel_filename, mel_spectrogram.T, allow_pickle=False)
#global condition features
if hparams.gin_channels > 0:
raise RuntimeError(
'When activating global conditions, please set your speaker_id rules in line 129 of datasets/wavenet_preprocessor.py to use them during training'
)
speaker_id = '<no_g>' #put the rule to determine how to assign speaker ids (using file names maybe? file basenames are available in "index" variable)
else:
speaker_id = '<no_g>'
# Return a tuple describing this training example
return (audio_filename, mel_filename, mel_filename, speaker_id, time_steps,
mel_frames)