This project is an AI-powered pipeline for automatically extracting potential highlight clips from Marvel Rivals gameplay footage. It uses audio-based spike detection to identify exciting moments (like shouting or hype reactions) and slices the video into short clips around those timestamps.

• 🔊 Audio Spike Detection: Uses RMS energy to find sudden increases in volume
• ✂️ Automatic Clip Extraction: Cuts 10-second clips around spike timestamps
• 📊 Waveform Plotting: Visualizes the audio signal with detected spikes
• 📂 Organized Output: Saves all clips into a highlights/ folder

Currently, the extracted clips are mostly triggered by voice activity, such as player reactions or commentary. These are not guaranteed gameplay highlights yet — future improvements will include frame-based analysis to better detect in-game events like kills, ultimates, or flashy moments.

highlight_extractor/
├── audio_spike_detector.py     # Detects audio spikes and saves timestamps
├── clip_extractor.py           # Cuts highlight clips based on spike timestamps
├── test_video.mp4              # Input gameplay video (replace with your own)
├── spike_timestamps.json       # JSON file with timestamps of detected spikes
├── highlights/                 # Output folder for extracted highlight clips
├── venv/                       # Virtual environment (excluded from Git)
└── .gitignore

• Python 3.10
• Librosa – audio analysis
• MoviePy – video slicing
• Matplotlib – plotting
• FFmpeg

• 🖼 Add frame-based highlight detection (OpenCV or YOLOv8)
• 🫣 Add Whisper transcription + GPT commentary (optional)
• 🎮 Merge highlights into a single reel
• 🧪 Build a CLI or web UI for usability

Brian Guida
GitHub: @brianguida