Anti-Video Forgery (AVF)

Detect video manipulations, deepfakes, and AI-generated content with our Anti-Video Forgery (AVF) service.

How to use?

The AVF service uses an asynchronous processing model:

1. Submit a video via the AVF Infer endpoint to start analysis
2. Receive a request ID in the response
3. Poll for results using the AVF Status endpoint with your request ID

How it works?

The AVF service analyses video content to detect signs of manipulation or AI generation:

1. Frame Analysis: Extracts frames from the submitted video
2. Temporal Analysis: Examines patterns across multiple frames (typically 50 frames)
3. Forgery Detection: Uses deep learning models to identify manipulation artifacts
4. Confidence Scoring: Returns a prediction with confidence score

Processing typically takes 1-2 minutes depending on video length.

Output Data

• prediction: Classification result
  • real - Video appears authentic
  • fake - Video shows signs of manipulation or AI generation
• confidence: Confidence level of the prediction (higher values = more confident)
• score: Numerical score indicating likelihood of forgery
• status: Processing status
  • pending - Request received, awaiting processing
  • processing - Video is being analysed
  • completed - Analysis complete
  • failed - Analysis failed

JSON Response Example

Successful Detection:

{
  "request_id": "246fcaf1-0c88-4b21-a96c-05cabefcef55",
  "status": "completed",
  "result": {
    "success": true,
    "message": "Inference completed successfully",
    "prediction": "fake",
    "confidence": 1.35,
    "score": 1.35,
    "metadata": {
      "sequence_length": 50,
      "video_frames": 134,
      "source_filename": "test.mp4"
    }
  }
}

Limitations

• Maximum file size: 100MB
• Processing timeout: 5 minutes
• Supported formats: MP4, AVI, MOV, and other common video formats

Use Cases

• Identity Verification: Detect deepfake attempts during video KYC
• Fraud Prevention: Identify manipulated video evidence
• Content Moderation: Flag AI-generated video content
• Security Screening: Verify authenticity of video submissions