Transcribe any audio/video file to text via OpenAI or offline Whisper backends. Auto-converts input with ffmpeg and outputs raw JSON/SRT/VTT or cleaned transcripts with timestamps and optional speaker labels/Markdown.