Automatically analyzes videos using AI to find the best 5 "meme-worthy" moments based on:
- Facial expressions and reactions
- Composition and sharpness
- Reaction quality (using CLIP AI model)
- Python script:
/scripts/meme_frame_picker.py - Python dependencies installed in
/venv - API endpoint:
/api/analyze-video - UI integrated in homepage
- When a video loads: The AI automatically starts analyzing it in the background
- Real-time display: As analysis completes, the top 5 frames appear below the video player
- Click to jump: Click any AI-selected frame to jump to that moment in the video
- One-click capture: Then hit "Merchify This" to capture that perfect frame
- AI-Powered Selection: Uses OpenAI's CLIP model to detect reaction faces
- Face Detection: Prioritizes frames with centered, visible faces
- Sharpness Filter: Only picks clear, high-quality frames
- Scored Results: Each frame gets a quality score (0-100%)
- Timeline Labels: Shows timestamp for each frame
- User opens homepage
- Selects a video
- AI analyzes in background (shows "Analyzing..." spinner)
- Top 5 frames appear below video
- User clicks a frame → video jumps to that moment
- User clicks "Merchify This" → captures frame
- Proceeds to customize page
- Python: OpenCV, PyTorch, OpenCLIP, NumPy
- Node.js API: Spawns Python process, handles results
- React UI: Real-time display with thumbnails
- Processes ~30s of video at a time
- Extracts 2 frames per second
- Analyzes in chunks for long videos
- Returns results incrementally
Ready to demo! 🚀