🎮 Auto Clip Extractor

An automated video processing tool that detects and extracts kill moments from Counter-Strike 2 gameplay recordings. The application uses computer vision techniques to identify kill feed events and automatically creates clip compilations.

📋 Table of Contents

• Features
• Installation
• Configuration
• How to Use
• Future Development

---

✨ Features

• Automatic Kill Detection: Uses template matching and edge detection to identify kill feed events
• Smart Clip Merging: Combines consecutive kills into single highlight clips
• ROI (Region of Interest): Focus detection on specific screen areas for better performance
• Color Filtering: Distinguishes between player kills (red border) and teammate kills (gray border)
• Multi-language Support: Available in Turkish and English
• Modern GUI: Built with CustomTkinter for a sleek, dark-themed interface
• Real-time Preview: Watch detection in action with live frame preview
• Batch Processing: Process multiple videos automatically
• Smart Caching: Skips already processed videos
• FFmpeg Integration: Fast, lossless clip extraction with audio preservation

---

🚀 Installation

Prerequisites

• Python 3.8 or higher
• FFmpeg (must be in system PATH)
• Windows OS (currently optimized for Windows)

Step 1: Install Python Dependencies

pip install opencv-python numpy customtkinter pillow

Or use the requirements file if provided:

pip install -r requirements.txt

Step 2: Install FFmpeg

1. Download FFmpeg from ffmpeg.org
2. Extract the archive
3. Add the bin folder to your system PATH
4. Verify installation: ffmpeg -version

Step 3: Prepare Template Image

You need a template image of the CS2 kill feed icon:

• Take a screenshot during a kill
• Crop just the kill feed icon/symbol
• Save as killfeed_template.jpg in the project root

Optional: For ROI preview, include an example.jpg screenshot showing your typical CS2 UI.

Step 4: Run the Application

Double-click start.bat or run:

pythonw gui.py

---

⚙️ Configuration

The application is configured through config.json. All settings can also be adjusted via the GUI Settings tab.

General Settings

Setting	Default	Description
LANGUAGE	"tr"	Interface language ("tr" or "en")
INPUT_FOLDER	"input_videos"	Folder containing videos to process
OUTPUT_FOLDER	"kills"	Folder where clips will be saved
TEMPLATE_PATH	"killfeed_template.jpg"	Path to kill feed template image

Detection Settings

Setting	Default	Description
THRESHOLD	0.55	Template matching threshold (0.0-1.0). Higher = more strict
FRAME_SKIP	120	Process every Nth frame (120 = every 2 seconds at 60fps)
KILL_COOLDOWN	2.0	Minimum seconds between kill detections
USE_EDGE_DETECTION	true	Use Canny edge detection for better accuracy
USE_COLOR_FILTER	true	Filter by red border (player kills only)
CANNY_THRESHOLD1	150	Canny edge detection lower threshold
CANNY_THRESHOLD2	250	Canny edge detection upper threshold
MIN_COLOR_PIXELS	150	Minimum red pixels to confirm player kill

Buffer Settings

Setting	Default	Description
BUFFER_BEFORE	4.0	Seconds to include before kill
BUFFER_AFTER	4.0	Seconds to include after kill
MIN_KILL_GAP	7.0	Merge kills within this time gap (seconds)

ROI (Region of Interest) Settings

Setting	Default	Description
USE_ROI	true	Only scan specific screen region
ROI_X_START	0.83	ROI left edge (0.0-1.0, percentage of width)
ROI_Y_START	0.07	ROI top edge (0.0-1.0, percentage of height)
ROI_X_END	1.0	ROI right edge (0.0-1.0, percentage of width)
ROI_Y_END	0.25	ROI bottom edge (0.0-1.0, percentage of height)

Color Filter Settings

Setting	Default	Description
KILL_COLOR_LOWER	[0, 120, 80]	HSV lower bound for red (range 1)
KILL_COLOR_UPPER	[10, 255, 255]	HSV upper bound for red (range 1)
KILL_COLOR_LOWER2	[170, 120, 80]	HSV lower bound for red (range 2)
KILL_COLOR_UPPER2	[180, 255, 255]	HSV upper bound for red (range 2)

---

📖 How to Use

1. Prepare Your Videos

1. Place your CS2 gameplay recordings in the input_videos folder
2. Supported formats: .mp4, .avi, .mov, .mkv, .flv, .wmv

2. Launch the Application

Run start.bat or execute pythonw gui.py

3. Navigate the Interface

🎬 Process Tab

• Click "▶ Start Processing" to begin
• Monitor real-time logs showing detection progress
• View live preview of detected kills
• Track progress bar and statistics

📁 Videos Tab

• View all videos in the input folder
• See processing status (✓ Processed / ⏳ Waiting)
• File size information
• Quick access to input folder

✂️ Clips Tab

• Browse all extracted clips
• See file details and source video
• Play clips directly from the interface
• Quick access to output folder

⚙️ Settings Tab

• Adjust all detection parameters
• Configure ROI with live preview overlay
• Change buffer times
• Modify detection thresholds
• Save settings with one click

4. Processing Flow

The application follows this workflow:

1. Scan Input: Discovers all video files in input_videos
2. Skip Processed: Checks processed_videos.json to avoid reprocessing
3. Detect Kills: Uses template matching to find kill feed events
4. Filter Results: Applies edge detection and color filtering
5. Merge Clips: Combines consecutive kills into single segments
6. Extract Clips: Uses FFmpeg to create video clips with buffers
7. Save Results: Outputs clips to kills folder with descriptive names

5. Output Format

Clips are named descriptively:

{original_video_name}_kill_{number}_{start_time}s-{end_time}s.mp4

Example:

competitive_match_kill_001_45.3s-52.1s.mp4

6. Tips for Best Results

• Template Quality: Use a clear, high-contrast screenshot of the kill feed icon
• ROI Configuration: Use the ROI preview to ensure the kill feed area is covered
• Threshold Tuning:
  • Too low (< 0.50): False positives
  • Too high (> 0.70): Missed kills
  • Start at 0.55 and adjust
• Frame Skip: Higher values = faster processing but may miss quick kills
• Color Filter: Disable if you want teammate kills included
• Buffer Times: Adjust for desired clip length (default: 8 seconds total)

---

🔮 Future Development

Planned Features

• AI-Powered Kill Detection: Machine learning models for more accurate and intelligent kill detection
• Multi-game Support: Expand to other FPS games (Valorant, Apex Legends, Call of Duty, Overwatch, etc.)
• More Detection Methods: Improved algorithms for better accuracy

---

📝 License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

🤝 Contributing

Contributions, issues, and feature requests are welcome! Feel free to check the issues page.

💡 Support

For issues, questions, or suggestions, please open an issue on the repository.

---

Made with ❤️ for the CS2 community