transkription

Transcribes audio/video files to text. Uses Whisper (openai-whisper) or Vosk (offline) as optional backend — both are detected via presence check. Without backend: placeholder mode with dummy output (dry-run).