Spaces:
Runtime error
Runtime error
Upload vits_pinyin.py
Browse files- vits_pinyin.py +88 -0
vits_pinyin.py
ADDED
@@ -0,0 +1,88 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import re
|
2 |
+
|
3 |
+
from pypinyin import Style
|
4 |
+
from pypinyin.contrib.neutral_tone import NeutralToneWith5Mixin
|
5 |
+
from pypinyin.converter import DefaultConverter
|
6 |
+
from pypinyin.core import Pinyin
|
7 |
+
|
8 |
+
from text import pinyin_dict
|
9 |
+
from bert import TTSProsody
|
10 |
+
|
11 |
+
|
12 |
+
class MyConverter(NeutralToneWith5Mixin, DefaultConverter):
|
13 |
+
pass
|
14 |
+
|
15 |
+
|
16 |
+
def is_chinese(uchar):
|
17 |
+
if uchar >= u'\u4e00' and uchar <= u'\u9fa5':
|
18 |
+
return True
|
19 |
+
else:
|
20 |
+
return False
|
21 |
+
|
22 |
+
|
23 |
+
def clean_chinese(text: str):
|
24 |
+
text = text.strip()
|
25 |
+
text_clean = []
|
26 |
+
for char in text:
|
27 |
+
if (is_chinese(char)):
|
28 |
+
text_clean.append(char)
|
29 |
+
else:
|
30 |
+
if len(text_clean) > 1 and is_chinese(text_clean[-1]):
|
31 |
+
text_clean.append(',')
|
32 |
+
text_clean = ''.join(text_clean).strip(',')
|
33 |
+
return text_clean
|
34 |
+
|
35 |
+
|
36 |
+
class VITS_PinYin:
|
37 |
+
def __init__(self, bert_path, device):
|
38 |
+
self.pinyin_parser = Pinyin(MyConverter())
|
39 |
+
self.prosody = TTSProsody(bert_path, device)
|
40 |
+
|
41 |
+
def get_phoneme4pinyin(self, pinyins):
|
42 |
+
result = []
|
43 |
+
count_phone = []
|
44 |
+
for pinyin in pinyins:
|
45 |
+
if pinyin[:-1] in pinyin_dict:
|
46 |
+
tone = pinyin[-1]
|
47 |
+
a = pinyin[:-1]
|
48 |
+
a1, a2 = pinyin_dict[a]
|
49 |
+
result += [a1, a2 + tone]
|
50 |
+
count_phone.append(2)
|
51 |
+
return result, count_phone
|
52 |
+
|
53 |
+
def chinese_to_phonemes(self, text):
|
54 |
+
text = clean_chinese(text)
|
55 |
+
phonemes = ["sil"]
|
56 |
+
chars = ['[PAD]']
|
57 |
+
count_phone = []
|
58 |
+
count_phone.append(1)
|
59 |
+
for subtext in text.split(","):
|
60 |
+
if (len(subtext) == 0):
|
61 |
+
continue
|
62 |
+
pinyins = self.correct_pinyin_tone3(subtext)
|
63 |
+
sub_p, sub_c = self.get_phoneme4pinyin(pinyins)
|
64 |
+
phonemes.extend(sub_p)
|
65 |
+
phonemes.append("sp")
|
66 |
+
count_phone.extend(sub_c)
|
67 |
+
count_phone.append(1)
|
68 |
+
chars.append(subtext)
|
69 |
+
chars.append(',')
|
70 |
+
phonemes.append("sil")
|
71 |
+
count_phone.append(1)
|
72 |
+
chars.append('[PAD]')
|
73 |
+
chars = "".join(chars)
|
74 |
+
char_embeds = self.prosody.get_char_embeds(chars)
|
75 |
+
char_embeds = self.prosody.expand_for_phone(char_embeds, count_phone)
|
76 |
+
return " ".join(phonemes), char_embeds
|
77 |
+
|
78 |
+
def correct_pinyin_tone3(self, text):
|
79 |
+
pinyin_list = [p[0] for p in self.pinyin_parser.pinyin(
|
80 |
+
text, style=Style.TONE3, strict=False, neutral_tone_with_five=True)]
|
81 |
+
if len(pinyin_list) >= 2:
|
82 |
+
for i in range(1, len(pinyin_list)):
|
83 |
+
try:
|
84 |
+
if re.findall(r'\d', pinyin_list[i-1])[0] == '3' and re.findall(r'\d', pinyin_list[i])[0] == '3':
|
85 |
+
pinyin_list[i-1] = pinyin_list[i-1].replace('3', '2')
|
86 |
+
except IndexError:
|
87 |
+
pass
|
88 |
+
return pinyin_list
|