Video Scene Detect MCP Server

Detect scene changes in videos and auto-generate YouTube chapters using advanced computer vision analysis.

Features

Tool	Description	Price
detect_scenes	Detect scene changes and return timestamps, frame numbers, and duration	$0.05/use
generate_chapters	Detect scenes and format as YouTube chapter timestamps (MM:SS Chapter)	$0.08/use

Connect via Claude Desktop

Add to your Claude Desktop MCP settings:

{
  "mcpServers": {
    "scene-detect": {
      "url": "https://ntriqpro--scene-detect-mcp.apify.actor/mcp?token=YOUR_APIFY_TOKEN"
    }
  }
}

Supported Video Formats

MP4, WebM, MKV, AVI, MOV, FLV, and other common video formats supported by FFmpeg.

Input

Both tools accept:

• video_url (required): URL of the video file to process
• threshold (optional): Scene detection sensitivity (0-100, default: 27.0)
  • Lower threshold = more scenes detected (more sensitive)
  • Higher threshold = fewer scenes detected (less sensitive)

Output Examples

detect_scenes

{
  "status": "success",
  "scenes": [
    {
      "scene_number": 1,
      "start_time": 0.0,
      "end_time": 15.5,
      "start_frame": 0,
      "end_frame": 465,
      "duration": 15.5
    },
    {
      "scene_number": 2,
      "start_time": 15.5,
      "end_time": 42.3,
      "start_frame": 465,
      "end_frame": 1269,
      "duration": 26.8
    }
  ],
  "total_scenes": 2,
  "video_duration": 42.3,
  "fps": 30
}

generate_chapters

{
  "status": "success",
  "chapters": "00:00 Scene 1\n00:15 Scene 2\n00:42 Scene 3",
  "total_scenes": 3,
  "video_duration": 120.5,
  "scenes_raw": [...]
}

Use Cases

1. Auto-Chapter Generation: Generate YouTube chapter timestamps for long-form videos
2. Content Segmentation: Automatically split videos into scenes for editing
3. Thumbnail Selection: Find key scene transitions for thumbnail generation
4. Video Analysis: Analyze scene composition and transitions
5. Accessibility: Generate chapter-based navigation for video players

Technology

• Scene Detection: PySceneDetect (BSD-3-Clause License)
• Video Processing: FFmpeg
• AI Backend: ntriq AI Server
• Framework: Model Context Protocol (MCP)

Open Source Licenses

This service uses the following open source projects:

• PySceneDetect — BSD-3-Clause License
• FFmpeg — LGPL/GPL License
• OpenAI Whisper — MIT License

Performance

• Processing Time: ~3-10 seconds per minute of video (depending on resolution and threshold)
• Maximum Duration: Up to 8 hours per video
• Resolution Support: 480p to 4K
• Accuracy: ~95% scene transition detection with default threshold

API Reference

detect_scenes

Analyzes a video file and returns all detected scene changes with precise timing information.

Parameters:

• video_url (string, required): Direct URL to video file
• threshold (number, optional): Detection sensitivity (default 27.0)

Returns:

• scenes: Array of scene objects with timing and frame data
• total_scenes: Count of scenes detected
• video_duration: Total video duration in seconds
• fps: Frames per second of the video

generate_chapters

Analyzes video scenes and generates YouTube-compatible chapter timestamps.

Parameters:

• video_url (string, required): Direct URL to video file
• threshold (number, optional): Detection sensitivity (default 27.0)

Returns:

• chapters: Newline-separated chapter timestamps (MM:SS format)
• total_scenes: Count of scenes detected
• video_duration: Total video duration
• scenes_raw: Full scene detection data

Rate Limiting

• 100 concurrent requests per API token
• 1,000 videos per day per token
• Contact for higher limits

Legal Disclaimer

Video Content Notice: Users are solely responsible for ensuring they have the rights to analyze and process videos through this service. This includes compliance with copyright laws, platform terms of service (YouTube, Vimeo, etc.), and any applicable regional content regulations. Videos are processed in real-time and are not stored or retained.

This service is provided as a tool for video analysis and does not constitute legal advice regarding video rights or intellectual property.

Support

For issues, questions, or feedback, contact the ntriq AI team at support@ntriq.co.kr

---

Platform: Apify Pricing Model: Pay-Per-Event Status: Production Ready