From 95b8866f3c40e3f94c74cb5c7bb28c839eb341c8 Mon Sep 17 00:00:00 2001 From: passerbya Date: Tue, 9 Jul 2024 17:25:55 +0800 Subject: [PATCH] =?UTF-8?q?=E4=BC=98=E5=85=88=E4=BD=BF=E7=94=A8ttsfrd?= =?UTF-8?q?=EF=BC=8Cttsfrd=E4=B8=8D=E5=AD=98=E5=9C=A8=E6=97=B6=E4=BD=BF?= =?UTF-8?q?=E7=94=A8WeTextProcessing?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- cosyvoice/cli/frontend.py | 25 +++++++++++++++++++++---- 1 file changed, 21 insertions(+), 4 deletions(-) diff --git a/cosyvoice/cli/frontend.py b/cosyvoice/cli/frontend.py index f10f655..c30135a 100644 --- a/cosyvoice/cli/frontend.py +++ b/cosyvoice/cli/frontend.py @@ -23,7 +23,12 @@ import os import inflect from tn.chinese.normalizer import Normalizer as ZhNormalizer from tn.english.normalizer import Normalizer as EnNormalizer - +try: + import ttsfrd + use_ttsfrd = True +except: + print("failed to import ttsfrd, please normalize input text manually") + use_ttsfrd = False from cosyvoice.utils.frontend_utils import contains_chinese, replace_blank, replace_corner_mark, remove_bracket, spell_out_number, split_paragraph @@ -50,8 +55,17 @@ class CosyVoiceFrontEnd: self.instruct = instruct self.allowed_special = allowed_special self.inflect_parser = inflect.engine() - self.zh_tn_model = ZhNormalizer(remove_erhua=False,full_to_half=False) - self.en_tn_model = EnNormalizer() + self.use_ttsfrd = use_ttsfrd + if self.use_ttsfrd: + self.frd = ttsfrd.TtsFrontendEngine() + ROOT_DIR = os.path.dirname(os.path.abspath(__file__)) + assert self.frd.initialize('{}/../../pretrained_models/CosyVoice-ttsfrd/resource'.format(ROOT_DIR)) is True, 'failed to initialize ttsfrd resource' + self.frd.set_lang_type('pinyin') + self.frd.enable_pinyin_mix(True) + self.frd.set_breakmodel_index(1) + else: + self.zh_tn_model = ZhNormalizer(remove_erhua=False,full_to_half=False) + self.en_tn_model = EnNormalizer() def _extract_text_token(self, text): text_token = self.tokenizer.encode(text, allowed_special=self.allowed_special) @@ -86,7 +100,10 @@ class CosyVoiceFrontEnd: def text_normalize(self, text, split=True): text = text.strip() if contains_chinese(text): - text = self.zh_tn_model.normalize(text) + if self.use_ttsfrd: + text = self.frd.get_frd_extra_info(text, 'input') + else: + text = self.zh_tn_model.normalize(text) text = text.replace("\n", "") text = replace_blank(text) text = replace_corner_mark(text)