forked from CSTR-Edinburgh/ophelia
-
Notifications
You must be signed in to change notification settings - Fork 8
/
synthesize_set_of_samples.py
82 lines (60 loc) · 2.85 KB
/
synthesize_set_of_samples.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
81
82
from configuration import load_config
from data_load import *
import numpy as np
from synthesize import *
from synthesize_with_latent_space import load
from tqdm import tqdm
import pandas as pd
from itertools import product
# conf_file='/home/noetits/doctorat_code/ophelia/config/blizzard_letters.cfg'
conf_file='./config/blizzard_unsupervised_letters.cfg'
# conf_file='/home/noetits/doctorat_code/ophelia/config/will_unsupervised_letters_unsup_graph_old_preprocess.cfg'
hp=load_config(conf_file)
model_type='unsup'
logdir = hp.logdir + "-" + model_type
##########################################
train_codes_pca=np.load(logdir+'/emo_codes_pca_train.npy')
min_xy=train_codes_pca.min(axis=0)
max_xy=train_codes_pca.max(axis=0)
pca_model=load(logdir)
a=np.mgrid[min_xy[0]:max_xy[0]:100j]
b=np.mgrid[min_xy[1]:max_xy[1]:100j]
X=np.array(list(product(a, b)))
codes=pca_model.inverse_transform(X)
def synthesize_set(hp, X, codes, inference_batch=1000):
tts=tts_model(hp, model_type=model_type)
# melfiles = ["{}/{}".format(hp.coarse_audio_dir, fname.replace("wav", "npy")) for fname in fnames]
texts=pd.read_csv('harvard_sentences.txt')[:5]
texts=[el.split('. ')[-1] for el in texts.iloc[:,0].tolist()]
idxs=np.arange(len(X)).tolist()
idxs = [str(i) for i in idxs]
for j,text in enumerate(texts):
ids=['sent_'+str(j)+'_code_'+s for s in idxs]
# tts.synthesize(text=[text]*len(X), emo_code=np.expand_dims(codes, axis=1), id=ids)
for b in range(int(len(X)/inference_batch)):
tts.synthesize(text=[text]*inference_batch, emo_code=np.expand_dims(codes[b*inference_batch:(b+1)*inference_batch], axis=1), id=ids[b*inference_batch:(b+1)*inference_batch])
rest_idx=len(X)-len(X)%inference_batch
tts.synthesize(text=[text]*(len(X)%inference_batch), emo_code=np.expand_dims(codes[rest_idx:], axis=1), id=ids[rest_idx:])
# for i in tqdm(range(len(X))):
# print('code no. ', i)
# code = np.array([np.array([codes[i]])])
# for j,text in texts.iterrows():
# print(text[0])
# sent=text[0].split('. ')[-1]
# print(sent)
# # code_str='_'.join(str(list(X[i]))[1:-1].split(', '))
# id='sent_'+str(j)+'_code_'+str(i)
# tts.synthesize(text=sent, emo_code=code, id=id)
synthesize_set(hp, X, codes, inference_batch=3000)
# from server.ice_tts_server import ICE_TTS_server
# ice=ICE_TTS_server(hp, X, codes, web_page='web_page_play_presynthesized_samples.html')
# from architectures import *
#g = Text2MelGraph(hp, mode="synthesize"); print("Graph 1 (t2m) loaded")
#g = Graph_style_unsupervised(hp, mode="train", load_in_memory=False)
#sess=tf.Session()
#sess.run(tf.global_variables_initializer())
#var_list = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES)
#code=extract_emo_code(hp, mels, g)
#print(code)
#import pdb
#pdb.set_trace()