whisper-local/whisper_local/transcriber.py

"""Whisper-Transkription via faster-whisper."""

import logging
import sys
from pathlib import Path

import numpy as np
from faster_whisper import WhisperModel

logger = logging.getLogger(__name__)


def _model_cache_dir() -> str | None:
    """Im gebündelten Modus: Modell neben der EXE cachen (portable).
    Im Entwicklungsmodus: None → HuggingFace-Standard-Cache."""
    if getattr(sys, "frozen", False):
        cache = Path(sys.executable).parent / "models"
        try:
            cache.mkdir(exist_ok=True)
            return str(cache)
        except OSError:
            return None  # Fallback auf HuggingFace-Standard-Cache
    return None


class Transcriber:
    def __init__(
        self,
        model_name: str = "small",
        compute_type: str = "int8",
        language: str = "de",
        model: WhisperModel | None = None,
    ):
        self.language = language
        if model is not None:
            self.model = model
        else:
            logger.info("Lade Whisper-Modell '%s' (compute_type=%s)...", model_name, compute_type)
            self.model = WhisperModel(
                model_name, compute_type=compute_type, download_root=_model_cache_dir()
            )
            logger.info("Modell geladen")

    def transcribe(self, audio: np.ndarray) -> str:
        """Transkribiert Audio-Array zu Text."""
        segments, _ = self.model.transcribe(audio, language=self.language)
        text = " ".join(segment.text.strip() for segment in segments if segment.text.strip())
        if text:
            logger.info("Transkribiert: %s", text)
        else:
            logger.info("Keine Sprache erkannt")
        return text