Dimitrius174 commited on
Commit
73b14c9
·
verified ·
1 Parent(s): c0096bc

Добавлены ударения (RUAccent) + нормализация

Browse files
Files changed (1) hide show
  1. eremin_tts.py +17 -1
eremin_tts.py CHANGED
@@ -31,6 +31,7 @@ import numpy as np
31
  import soundfile as sf
32
  import torch
33
  import torchaudio
 
34
 
35
  from f5_tts.infer.utils_infer import (
36
  infer_process,
@@ -142,11 +143,26 @@ def main():
142
  model, vocoder = load_f5_model()
143
  ref_audio_p, ref_text_p = prepare_ref()
144
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
145
  raw_sents = [s.strip() for s in SENT_SPLIT_RE.split(args.text)
146
  if s and s.strip()]
147
  if not raw_sents:
148
  raw_sents = [args.text]
149
- sentences = [normalize_numbers(s) for s in raw_sents]
150
 
151
  print(f"Предложений: {len(sentences)}")
152
  for i, s in enumerate(sentences, 1):
 
31
  import soundfile as sf
32
  import torch
33
  import torchaudio
34
+ from ruaccent import RUAccent
35
 
36
  from f5_tts.infer.utils_infer import (
37
  infer_process,
 
143
  model, vocoder = load_f5_model()
144
  ref_audio_p, ref_text_p = prepare_ref()
145
 
146
+ # RUAccent: расстановка ударений
147
+ print("Loading RUAccent...")
148
+ accentizer = RUAccent()
149
+ accentizer.load(omograph_model_size='turbo3.1', use_dictionary=True,
150
+ tiny_mode=False)
151
+ print(" RUAccent loaded.")
152
+
153
+ def _accent(text):
154
+ if not text.strip() or '+' in text:
155
+ return text
156
+ try:
157
+ return accentizer.process_all(text)
158
+ except Exception:
159
+ return text
160
+
161
  raw_sents = [s.strip() for s in SENT_SPLIT_RE.split(args.text)
162
  if s and s.strip()]
163
  if not raw_sents:
164
  raw_sents = [args.text]
165
+ sentences = [_accent(normalize_numbers(s)) for s in raw_sents]
166
 
167
  print(f"Предложений: {len(sentences)}")
168
  for i, s in enumerate(sentences, 1):