Can I vibecode Swell AI?

Build a personal replacement for Swell AI in an empty repository.
Use Python 3.12, FastAPI, ffmpeg (via python-ffmpeg), SQLite, HTMX, and Tailwind CSS; do not offer alternative stacks.

DATA MODEL:
- `Episode`: id, original_filename, original_filepath, status (uploading, processing, ready, failed), created_at
- `Transcript`: id, episode_id, raw_json, formatted_text, srt_content, vtt_content, created_at
- `ContentItem`: id, episode_id, item_type (show_notes, blog_post, twitter_thread, linkedin_post), content, created_at
- `Clip`: id, episode_id, start_time_ms, end_time_ms, text, status

USER JOURNEYS & SCREENS:
1. **Dashboard (`/`)**: A clean grid showing all uploaded `Episodes`. Include a drag-and-drop zone at the top for uploading new audio/video files (MP3, WAV, MP4). Use a progress bar for active uploads.
2. **Episode Detail (`/episode/{id}`)**: A split-screen interface. Left side: Audio player with a generated waveform (use wavesurfer.js) and the interactive transcript (clicking text seeks audio). Right side: Tabs for 'Content', 'Clips', and 'Export'.
3. **Content Generation (`/episode/{id}/content`)**: Buttons to trigger generation of Show Notes, Blog Posts, and Social Threads. Display the generated text in editable textareas.
4. **Clip Editor (`/episode/{id}/clips`)**: An interface to highlight text in the transcript and click "Create Clip", which defines a start/end time. Display a list of saved clips.

TECHNICAL SPECIFICATIONS:
- Audio Processing: Use `ffmpeg` to extract audio from video and normalize loudness to -16 LUFS. Implement silence trimming and fades.
- Transcription: Integrate with OpenAI's Whisper API (`whisper-1` model) via the official Python SDK. Request word-level timestamps (`timestamp_granularities=["word"]`).
- Content Generation: Use Anthropic's Claude 3.5 Sonnet API for text generation. Prompt it with the full transcript to generate highly accurate show notes, blog posts, and viral tweets.
- Background Jobs: Since transcription and generation take time, use Python's `asyncio` and `BackgroundTasks` in FastAPI, updating the UI via HTMX polling (`hx-trigger="every 2s"`).
- Storage: Save uploaded media in a local `./media/uploads/` directory. Save exports in `./media/exports/`.

Make the first run work locally with one documented command.
Store all user data locally by default.
Put API keys (OPENAI_API_KEY, ANTHROPIC_API_KEY) in .env, ship .env.example.
Include clear empty, loading, success, and recoverable error states (e.g., API failures).
Write focused tests for the core processing pipeline and one end-to-end happy path.
Create a README with setup, architecture, and backup steps.
Do not add accounts, billing, telemetry, analytics, or a hosted control plane.