🔴 Phase 3: Audio-to-Text with Whisper Fallback

## Phase 3: Advanced Content Processing

Handle videos without captions using Whisper transcription.

### Tasks
- [ ] Integrate OpenAI Whisper API / local Whisper model
- [ ] Build audio extraction pipeline from YouTube videos
- [ ] Implement automatic fallback: captions → Whisper
- [ ] Add job queue for async transcription (Celery/RQ)
- [ ] Store generated transcripts alongside extracted ones
- [ ] Add progress tracking for transcription jobs

### Acceptance Criteria
- System automatically detects videos without captions
- Whisper transcribes audio to text
- Transcription jobs run asynchronously
- Users notified when transcription completes
- Supports multiple languages (if feasible)

### Tech Stack
- OpenAI Whisper (cloud or local)
- yt-dlp for audio extraction
- Celery/Redis Queue for async jobs
- FFmpeg for audio processing