forked from Rayhane-mamah/Tacotron-2
-
Notifications
You must be signed in to change notification settings - Fork 1
/
synthesize.py
80 lines (63 loc) · 3.3 KB
/
synthesize.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
import argparse
import os
from warnings import warn
from time import sleep
import tensorflow as tf
from hparams import hparams
from infolog import log
from tacotron.synthesize import tacotron_synthesize
def prepare_run(args):
modified_hp = hparams.parse(args.hparams)
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
run_name = args.name or args.tacotron_name or args.model
taco_checkpoint = os.path.join('logs-' + run_name, 'taco_' + args.checkpoint)
return taco_checkpoint, modified_hp
def get_sentences(args):
if args.text_list != '':
with open(args.text_list, encoding='utf-8') as f:
sentences = list(map(lambda l: l.strip('\t\r\n'), f.readlines()))
else:
sentences = hparams.sentences
texts=[]
speakers=[]
languages=[]
for i, line in enumerate(sentences):
line = line.split('|')
if len(line) == 3:
texts.append(line[0])
speakers.append(int(line[1]))
languages.append(int(line[2])) #0: English, 1: Chinese
return texts, speakers, languages
def main():
accepted_modes = ['eval', 'synthesis', 'live']
parser = argparse.ArgumentParser()
parser.add_argument('--checkpoint', default='pretrained/', help='Path to model checkpoint')
parser.add_argument('--hparams', default='',
help='Hyperparameter overrides as a comma-separated list of name=value pairs')
parser.add_argument('--name', help='Name of logging directory if the two models were trained together.')
parser.add_argument('--tacotron_name', help='Name of logging directory of Tacotron. If trained separately')
parser.add_argument('--wavenet_name', help='Name of logging directory of WaveNet. If trained separately')
parser.add_argument('--model', default='Tacotron')
parser.add_argument('--input_dir', default='training_data/', help='folder to contain inputs sentences/targets')
parser.add_argument('--mels_dir', default='tacotron_output/eval/', help='folder to contain mels to synthesize audio from using the Wavenet')
parser.add_argument('--output_dir', default='output/', help='folder to contain synthesized mel spectrograms')
parser.add_argument('--mode', default='eval', help='mode of run: can be one of {}'.format(accepted_modes))
parser.add_argument('--GTA', default='True', help='Ground truth aligned synthesis, defaults to True, only considered in synthesis mode')
parser.add_argument('--text_list', default='', help='Text file contains list of texts to be synthesized. Valid if mode=eval')
parser.add_argument('--speaker_id', default=None, help='Defines the speakers ids to use when running standalone Wavenet on a folder of mels. this variable must be a comma-separated list of ids')
args = parser.parse_args()
accepted_models = ['Tacotron']
if args.model not in accepted_models:
raise ValueError('please enter a valid model to synthesize with: {}'.format(accepted_models))
if args.mode not in accepted_modes:
raise ValueError('accepted modes are: {}, found {}'.format(accepted_modes, args.mode))
if args.GTA not in ('True', 'False'):
raise ValueError('GTA option must be either True or False')
taco_checkpoint, hparams = prepare_run(args)
sentences, speakers, languages = get_sentences(args)
if args.model == 'Tacotron':
_ = tacotron_synthesize(args, hparams, taco_checkpoint, sentences, speakers, languages)
else:
raise ValueError('Model provided {} unknown! {}'.format(args.model, accepted_models))
if __name__ == '__main__':
main()