59 lines
2.3 KiB
Python
59 lines
2.3 KiB
Python
|
import os
|
||
|
|
import logging
|
||
|
|
import asyncio
|
||
|
|
from concurrent.futures import ThreadPoolExecutor
|
||
|
|
from typing import Optional
|
||
|
|
|
||
|
|
logger = logging.getLogger('Superviseur')
|
||
|
|
|
||
|
|
class WhisperWorker:
|
||
|
|
"""Handles local STT transcription using faster-whisper."""
|
||
|
|
|
||
|
|
def __init__(self, model_size="large-v3", device="cpu"):
|
||
|
|
self.model_size = model_size
|
||
|
|
self.device = device
|
||
|
|
self.model = None
|
||
|
|
self.executor = ThreadPoolExecutor(max_workers=2)
|
||
|
|
self.initialized = False
|
||
|
|
|
||
|
|
def initialize(self):
|
||
|
|
"""Load the model into RAM/VRAM."""
|
||
|
|
try:
|
||
|
|
from faster_whisper import WhisperModel
|
||
|
|
logger.info(f"◈ SYSTEM: Initializing Whisper ({self.model_size}) on {self.device}...")
|
||
|
|
|
||
|
|
# On tente le chargement. Si device="cuda" échoue (AMD), on bascule sur CPU.
|
||
|
|
try:
|
||
|
|
self.model = WhisperModel(self.model_size, device=self.device, compute_type="int8")
|
||
|
|
except Exception as e:
|
||
|
|
if self.device != "cpu":
|
||
|
|
logger.warning(f"◈ GPU failure ({e}). Falling back to CPU.")
|
||
|
|
self.device = "cpu"
|
||
|
|
self.model = WhisperModel(self.model_size, device="cpu", compute_type="int8")
|
||
|
|
else:
|
||
|
|
raise e
|
||
|
|
|
||
|
|
self.initialized = True
|
||
|
|
logger.info(f"◈ SYSTEM: Whisper Model Loaded on {self.device}. Ear Protocol Active.")
|
||
|
|
except Exception as e:
|
||
|
|
logger.error(f"Failed to initialize Whisper: {e}")
|
||
|
|
self.initialized = False
|
||
|
|
|
||
|
|
async def transcribe(self, audio_path: str) -> Optional[str]:
|
||
|
|
"""Transcribe an audio file asynchronously."""
|
||
|
|
if not self.initialized or not self.model:
|
||
|
|
return None
|
||
|
|
|
||
|
|
def _run():
|
||
|
|
try:
|
||
|
|
segments, info = self.model.transcribe(audio_path, beam_size=5)
|
||
|
|
text = " ".join([segment.text for segment in segments]).strip()
|
||
|
|
logger.debug(f"◈ Whisper: Transcription result: '{text}' (lang: {info.language})")
|
||
|
|
return text
|
||
|
|
except Exception as e:
|
||
|
|
logger.error(f"Transcription error: {e}")
|
||
|
|
return None
|
||
|
|
|
||
|
|
loop = asyncio.get_event_loop()
|
||
|
|
return await loop.run_in_executor(self.executor, _run)
|