"""Adapter wrapping faster-whisper into a small transcriber class. Provides a `FasterWhisperTranscriber` with a stable `transcribe` API that other code can depend on. Uses the implementation in `whisper_project.infra.transcribe`. """ from typing import Optional from whisper_project.infra.transcribe import transcribe_faster_whisper, write_srt class FasterWhisperTranscriber: def __init__(self, model: str = "base", compute_type: str = "int8") -> None: self.model = model self.compute_type = compute_type def transcribe(self, file_path: str, srt_out: Optional[str] = None): """Transcribe the given audio file. If `srt_out` is provided, writes an SRT file using `write_srt`. Returns the segments list (as returned by faster-whisper wrapper). """ segments = transcribe_faster_whisper(file_path, self.model, compute_type=self.compute_type) if srt_out and segments: write_srt(segments, srt_out) return segments __all__ = ["FasterWhisperTranscriber"] from typing import List from ..core.models import Segment class FasterWhisperTranscriber: """Adaptador que usa faster-whisper para transcribir y escribir SRT.""" def __init__(self, model: str = "base", compute_type: str = "int8"): self.model = model self.compute_type = compute_type def transcribe(self, audio_path: str, srt_out: str) -> List[Segment]: # Importar localmente para evitar coste al importar el módulo from faster_whisper import WhisperModel from whisper_project.transcribe import write_srt, dedupe_adjacent_segments model_obj = WhisperModel(self.model, device="cpu", compute_type=self.compute_type) segments_gen, info = model_obj.transcribe(audio_path, beam_size=5) segments = list(segments_gen) # Convertir a nuestros Segment dataclass result_segments = [] for s in segments: # faster-whisper segment tiene .start, .end, .text seg = Segment(start=float(s.start), end=float(s.end), text=str(s.text)) result_segments.append(seg) # escribir SRT usando la función existente (acepta objetos con .start/.end/.text) segments_to_write = dedupe_adjacent_segments(result_segments) write_srt(segments_to_write, srt_out) return result_segments