The Scene Splitting extractor accepts a video input and returns multiple documents split by silence, scene or time.

​
Overview

The Scene Splitting feature extractor analyzes video content to detect and segment distinct scenes based on visual changes. It breaks down videos into logical segments that can be processed and searched independently.

​
Required Inputs

ParameterTypeRequiredDefaultDescription
video_urlstringYes-URL pointing to the video file to be processed. Supported formats: MP4, MOV, AVI
min_scene_durationfloatNo1.0Minimum duration (in seconds) for a scene segment
thresholdfloatNo0.3Sensitivity threshold for scene change detection (0.0-1.0)

​
Configurations

​
Split Types

When processing video content, the extractor supports three different split methods:

Split TypeDescriptionCommon Use Cases
sceneSplits video based on visual content changesContent analysis, scene detection
silenceSplits video at periods of audio silenceSpeech segmentation, presentation slides
timeSplits video at fixed time intervalsRegular sampling, uniform segments

​
Configuration Examples

Scene-based
{
  "split_type": "scene",
  "threshold": 0.35,           // Scene change sensitivity (0-1)
  "min_scene_length": 2.5,     // Minimum scene duration in seconds
  "max_scene_length": 60,      // Maximum scene duration in seconds
  "keyframe_method": "content" // How to select the keyframe
}
Silence-based
{
  "split_type": "silence",
  "silence_threshold": -30,    // dB threshold for silence detection
  "min_silence_length": 1.0,   // Minimum silence duration
  "padding": 0.1              // Padding around silence boundaries
}
Time-based
{
  "split_type": "time",
  "interval": 10.0,           // Fixed interval in seconds
  "overlap": 0.0              // Overlap between segments
}

​
Base Configuration

OptionTypeDefaultDescription
min_scene_lengthfloat2.0Minimum scene duration in seconds
max_scene_lengthfloat120.0Maximum scene duration in seconds (will force-split longer scenes)

​
Configuration Examples

Sample
{
  "min_scene_length": 2.5,     // Minimum scene duration in seconds
  "max_scene_length": 60,      // Maximum scene duration in seconds
}

​
Processing Flow

​
Output Schema

This feature extractor will output as multiple documents.

{
  "document_id": "doc_abc123",
  "collection_id": "col_xyz789",
  "source_object_id": "obj_def456",
  
  // Scene metadata
  "scene": {
    "index": 3,               // Scene index in the video (0-based)
    "start_time": 42.5,       // Start timestamp in seconds
    "end_time": 68.2,         // End timestamp in seconds
    "duration": 25.7,         // Scene duration in seconds
    "keyframe_time": 55.2,    // Timestamp of the keyframe
    "keyframe_url": "https://storage.example.com/keyframes/abc123.jpg",
    "confidence": 0.87        // Confidence score of scene boundary detection
  },
  
  // Video metadata
  "video": {
    "filename": "company_event_2023.mp4",
    "total_duration": 1258.4,  // Total video duration in seconds
    "total_scenes": 17,        // Total number of detected scenes
    "width": 1920,
    "height": 1080,
    "fps": 30
  }
}

Was this page helpful?

Suggest editsRaise issue