def preprocess(args, input_folders, out_dir, hparams):
mel_dir = os.path.join(out_dir, 'mels')
wav_dir = os.path.join(out_dir, 'audio')
linear_dir = os.path.join(out_dir, 'linear')
os.makedirs(mel_dir, exist_ok=True)
os.makedirs(wav_dir, exist_ok=True)
os.makedirs(linear_dir, exist_ok=True)
if args.dataset.startswith('liepa'):
metadata = preprocessor_liepa.build_from_path(hparams,
input_folders,
None,
mel_dir,
linear_dir,
wav_dir,
args.n_jobs,
tqdm=tqdm)
else:
metadata = preprocessor.build_from_path(hparams,
input_folders,
mel_dir,
linear_dir,
wav_dir,
args.n_jobs,
tqdm=tqdm)
write_metadata(metadata, out_dir)
def preprocess(args, input_folders, output_dir, hparams):
mel_frames, timesteps = 0, 0
max_text_lens, max_mel_lens, max_timestep_lens = [], [], []
for input_dir in input_folders:
wav_dir = os.path.join(output_dir, input_dir.split('/')[-1], 'audio')
mel_dir = os.path.join(output_dir, input_dir.split('/')[-1], 'mels')
os.makedirs(wav_dir, exist_ok=True)
os.makedirs(mel_dir, exist_ok=True)
metadata = preprocessor.build_from_path(hparams,
input_dir,
wav_dir,
mel_dir,
args.n_jobs,
tqdm=tqdm)
with open(
os.path.join(output_dir,
input_dir.split('/')[-1], 'train.txt'), 'w') as f:
for m in metadata:
f.write('|'.join([str(x) for x in m]) + '\n')
max_text_lens.append(max(len(m[3]) for m in metadata))
max_mel_lens.append(max(int(m[2]) for m in metadata))
max_timestep_lens.append(max(m[1] for m in metadata))
mel_frames += sum([int(m[2]) for m in metadata])
timesteps += sum([int(m[1]) for m in metadata])
hours = timesteps / hparams.sample_rate / 3600
print(
f'Write {len(metadata)} utterances, {mel_frames} mel frames, {timesteps} audio timesteps, ({hours:.2f} hours)'
)
print(f'Max input length (text chars): {max(max_text_lens)}')
print(f'Max mel frames length: {max(max_mel_lens)}')
print(f'Max audio timesteps length: {max(max_timestep_lens)}')
def preprocess(args, input_folders, out_dir):
input_dir = os.path.join(out_dir, 'inputs')
label_dir = os.path.join(out_dir, 'labels')
os.makedirs(input_dir, exist_ok=True)
os.makedirs(label_dir, exist_ok=True)
metadata = preprocessor.build_from_path(input_folders, input_dir, label_dir, args, tqdm=tqdm)
write_metadata(metadata, out_dir, args)
def preprocess(args, input_folder, out_dir, hparams):
os.makedirs(out_dir, exist_ok=True)
metadata = preprocessor.build_from_path(hparams,
input_folder,
out_dir,
args.n_jobs,
tqdm=tqdm)
write_metadata(metadata, out_dir)
def preprocess(args):
in_dir = os.path.join(args.base_dir, args.input)
out_dir = os.path.join(args.base_dir, args.output)
os.makedirs(out_dir, exist_ok=True)
metadata = preprocessor.build_from_path(in_dir,
out_dir,
args.n_jobs,
tqdm=tqdm)
write_metadata(metadata, out_dir)
def preprocess(args, input_folders, out_dir, hparams):
mel_dir = os.path.join(out_dir, 'mels')
os.makedirs(mel_dir, exist_ok=True)
metadata = preprocessor.build_from_path(hparams,
input_folders,
mel_dir,
args.n_jobs,
tqdm=tqdm)
write_metadata(metadata, out_dir, hparams)
def preprocess(args, input_folders, out_dir, hparams): mel_dir = os.path.join(out_dir, 'mels') wav_dir = os.path.join(out_dir, 'audio') linear_dir = os.path.join(out_dir, 'linear') os.makedirs(mel_dir, exist_ok=True) os.makedirs(wav_dir, exist_ok=True) os.makedirs(linear_dir, exist_ok=True) metadata = preprocessor.build_from_path(hparams, input_folders, mel_dir, linear_dir, wav_dir, args.n_jobs, tqdm=tqdm) write_metadata(metadata, out_dir)
def preprocess(args, input_folders, out_dir): mel_dir = os.path.join(out_dir, 'mels') wav_dir = os.path.join(out_dir, 'audio') linear_dir = os.path.join(out_dir, 'linear') os.makedirs(mel_dir, exist_ok=True) os.makedirs(wav_dir, exist_ok=True) os.makedirs(linear_dir, exist_ok=True) metadata = preprocessor.build_from_path(input_folders, mel_dir, linear_dir, wav_dir, args.n_jobs, tqdm=tqdm) write_metadata(metadata, out_dir)
def preprocess(args, input_folder, out_dir, hparams):
cmp_dir = os.path.join(out_dir, 'cmp')
linear_dir = os.path.join(out_dir, 'linear')
for d in [cmp_dir, linear_dir]:
#if(os.path.exists(d)):
# shutil.rmtree(d)
os.makedirs(d, exist_ok=True)
metadata = preprocessor.build_from_path(hparams,
input_folder,
cmp_dir,
linear_dir,
args.n_jobs,
tqdm=tqdm)
write_metadata(metadata, out_dir)
def preprocess(args, input_folders, out_dir, hparams):
mel_dir = os.path.join(out_dir, 'mels')
wav_dir = os.path.join(out_dir, 'audio')
linear_dir = os.path.join(out_dir, 'linear')
os.makedirs(mel_dir, exist_ok=True)
os.makedirs(wav_dir, exist_ok=True)
os.makedirs(linear_dir, exist_ok=True)
speaker_wavs = get_LibriTTS_wavs()
metadata = preprocessor.build_from_path(hparams,
input_folders,
mel_dir,
linear_dir,
wav_dir,
args.n_jobs,
speaker_wavs,
tqdm=tqdm)
write_metadata(metadata, out_dir)
def preprocess(args, input_folders, out_dir, hparams):
mel_dir = os.path.join(out_dir, 'mels')
wav_dir = os.path.join(out_dir, 'audio')
linear_dir = os.path.join(out_dir, 'linear')
if not os.path.exists(mel_dir):
os.makedirs(mel_dir)
if not os.path.exists(wav_dir):
os.makedirs(wav_dir)
if not os.path.exists(linear_dir):
os.makedirs(linear_dir)
metadata = preprocessor.build_from_path(hparams,
input_folders,
mel_dir,
linear_dir,
wav_dir,
args.n_jobs,
tqdm=tqdm)
write_metadata(metadata, out_dir)
def preprocess(args, hparams):
input_dir = norm_data(args)
output_dir = os.path.join(args.base_dir, 'training_data')
mel_dir = os.path.join(output_dir, 'mels')
wav_dir = os.path.join(output_dir, 'audio')
linear_dir = os.path.join(output_dir, 'linear')
os.makedirs(mel_dir, exist_ok=True)
os.makedirs(wav_dir, exist_ok=True)
os.makedirs(linear_dir, exist_ok=True)
metadata = preprocessor.build_from_path(hparams,
input_dir,
mel_dir,
linear_dir,
wav_dir,
args.n_jobs,
tqdm=tqdm)
write_metadata(metadata, output_dir)
def preprocess(args, wav_folder, out_dir, hparams):
in_dir = os.path.join(args.folder_wav_dir, wav_folder)
mel_dir = os.path.join(out_dir, 'mels')
audio_dir = os.path.join(out_dir, 'audio')
linear_dir = os.path.join(out_dir, 'linear')
spk_emb_dir = os.path.join(out_dir, 'spkemb')
os.makedirs(mel_dir, exist_ok=True)
os.makedirs(audio_dir, exist_ok=True)
os.makedirs(linear_dir, exist_ok=True)
os.makedirs(spk_emb_dir, exist_ok=True)
# metadata = preprocessor.build_from_path(hparams, mel_dir, linear_dir, wav_dir, args.n_jobs, tqdm=tqdm)
metadata = preprocessor.build_from_path(hparams,
args,
in_dir,
mel_dir,
linear_dir,
audio_dir,
spk_emb_dir,
args.n_jobs,
tqdm=tqdm)
write_metadata(metadata, out_dir)
def preprocess(args, input_folders, out_dir, hparams):
mel_dir = os.path.join(out_dir, 'mels')
wav_dir = os.path.join(out_dir, 'audio')
linear_dir = os.path.join(out_dir, 'linear')
os.makedirs(mel_dir, exist_ok=True)
os.makedirs(wav_dir, exist_ok=True)
os.makedirs(linear_dir, exist_ok=True)
if args.dataset.startswith("VCTK"):
metadata, speaker_dict = preprocessor_VCTK.build_from_path(
hparams,
input_folders,
mel_dir,
linear_dir,
wav_dir,
args.n_jobs,
tqdm=tqdm)
pickle.dump(speaker_dict,
open(os.path.join(out_dir, "speaker_dict.pkl"), "wb"))
elif args.dataset.startswith("SynPaFlex"):
metadata, speaker_dict = preprocessor_SynPaFlex.build_from_path(
hparams,
input_folders,
mel_dir,
linear_dir,
wav_dir,
args.n_jobs,
tqdm=tqdm)
pickle.dump(speaker_dict,
open(os.path.join(out_dir, "speaker_dict.pkl"), "wb"))
else:
metadata = preprocessor.build_from_path(hparams,
input_folders,
mel_dir,
linear_dir,
wav_dir,
args.n_jobs,
tqdm=tqdm)
write_metadata(metadata, out_dir)
def preprocess(args, input_folders, out_dir, hparams):
mel_dir = os.path.join(out_dir, 'mels')
wav_dir = os.path.join(out_dir, 'audio')
linear_dir = os.path.join(out_dir, 'linear')
os.makedirs(mel_dir, exist_ok=True)
os.makedirs(wav_dir, exist_ok=True)
os.makedirs(linear_dir, exist_ok=True)
if args.dataset == 'KRSPEECH':
metadata = krspeech.build_from_path(hparams,
input_folders,
mel_dir,
linear_dir,
wav_dir,
args.n_jobs,
tqdm=tqdm)
else:
metadata = preprocessor.build_from_path(hparams,
input_folders,
mel_dir,
linear_dir,
wav_dir,
args.n_jobs,
tqdm=tqdm)
write_metadata(metadata, out_dir)
