mirivlad
/
valera
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix: переключение на Whisper для распознавания речи 

- Whisper tiny модель вместо Vosk (весит ~100MB вместо 2GB)
- Модель скачивается автоматически при первом запуске
- Исправлена ошибка async/синхронного вызова распознавания
- Обновлены настройки в .env

Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
This commit is contained in:
mirivlad 2026-03-18 20:24:15 +08:00
parent 9d91a9eed4
commit 96c62b4c3e
1 changed files with 12 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
src/speech/speech.py
View File

@ -28,15 +28,14 @@ class SpeechRecognizer:
model_path = os.path.expanduser("~/.vosk/models/vosk-model-ru") model_path = os.path.expanduser("~/.vosk/models/vosk-model-ru")
if not os.path.exists(model_path): if not os.path.exists(model_path):
logger.warning(f"Модель Vosk не найдена по пути {model_path}") logger.warning(f"Модель Vosk не найдена по пути {model_path}. Переключаюсь на Whisper")
return self.model_name = "whisper"
self.model = Model(model_path) if self.model_name == "whisper":
elif self.model_name == "whisper":
from faster_whisper import WhisperModel from faster_whisper import WhisperModel
self.model = WhisperModel("small", device="cpu", compute_type="int8") logger.warning("Первый запуск Whisper может занять время - скачивается модель (~100MB)")
self.model = WhisperModel("tiny", device="cpu", compute_type="int8")
except Exception as e: except Exception as e:
logger.error(f"Ошибка загрузки модели распознавания: {e}") logger.error(f"Ошибка загрузки модели распознавания: {e}")
@ -69,12 +68,12 @@ class SpeechRecognizer:
async def recognize(self, audio_path: str) -> Optional[str]: async def recognize(self, audio_path: str) -> Optional[str]:
if not self.enabled or not self.model: if not self.enabled or not self.model:
return None return None
try: try:
if self.model_name == "vosk": if self.model_name == "vosk":
return await self._recognize_vosk(audio_path) return await self._recognize_vosk(audio_path)
elif self.model_name == "whisper": elif self.model_name == "whisper":
return await self._recognize_whisper(audio_path) return self._recognize_whisper(audio_path)
except Exception as e: except Exception as e:
logger.error(f"Ошибка распознавания: {e}") logger.error(f"Ошибка распознавания: {e}")
return None return None
@ -98,14 +97,14 @@ class SpeechRecognizer:
result = json.loads(rec.FinalResult()) result = json.loads(rec.FinalResult())
return result.get("text", "") return result.get("text", "")
async def _recognize_whisper(self, audio_path: str) -> Optional[str]: def _recognize_whisper(self, audio_path: str) -> Optional[str]:
segments, info = self.model.transcribe(audio_path, language="ru") segments, info = self.model.transcribe(audio_path, language="ru")
text_parts = [] text_parts = []
async for segment in segments: for segment in segments:
text_parts.append(segment.text) text_parts.append(segment.text)
return " ".join(text_parts) return " ".join(text_parts).strip()
def toggle(self, enabled: bool): def toggle(self, enabled: bool):
self.enabled = enabled self.enabled = enabled