Skip to content

SOTA Open Source TTS

replace cn2an with WeTextProcessing for text normalization (#33)

faceair597970a1a53a
diff --git a/fish_speech/text/chinese.py b/fish_speech/text/chinese.py
index 3cdec82..2133a9a 100644
--- a/fish_speech/text/chinese.py
+++ b/fish_speech/text/chinese.py
@@ -1,9 +1,9 @@
import os
import re
-import cn2an
import jieba.posseg as psg
from pypinyin import Style, lazy_pinyin
+from tn.chinese.normalizer import Normalizer
from fish_speech.text.symbols import punctuation
from fish_speech.text.tone_sandhi import ToneSandhi
@@ -16,7 +16,7 @@ pinyin_to_symbol_map = {
for line in open(OPENCPOP_DICT_PATH).readlines()
}
-
+normalizer = Normalizer()
tone_modifier = ToneSandhi()
@@ -123,10 +123,7 @@ def _g2p(segments):
def text_normalize(text):
- numbers = re.findall(r"\d+(?:\.?\d+)?", text)
- for number in numbers:
- text = text.replace(number, cn2an.an2cn(number), 1)
- return text
+ return normalizer.normalize(text)
if __name__ == "__main__":
diff --git a/pyproject.toml b/pyproject.toml
index 18c4073..eca2885 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -24,7 +24,6 @@ dependencies = [
"vector-quantize-pytorch>=1.10.0",
"rich>=13.5.3",
"gradio>=4.0.0",
- "cn2an>=0.5.22",
"pypinyin>=0.49.0",
"jieba>=0.42.1",
"g2p-en>=2.1.0",
@@ -34,6 +33,7 @@ dependencies = [
"kui>=1.6.0",
"zibai-server>=0.9.0",
"loguru>=0.6.0",
+ "WeTextProcessing>=0.1.10",
]
[project.optional-dependencies]