Video Detection
Video Detection
Section titled “Video Detection”Manuscript analyzes video files to detect deepfakes, face swaps, and fully AI-generated video content.
Detection Signals
Section titled “Detection Signals”Container Metadata
Section titled “Container Metadata”Video containers contain rich metadata:
| Metadata | Real Video | AI-Generated |
|---|---|---|
| Encoder | Standard codecs (H.264, H.265) | May use non-standard |
| Creation software | Premiere, Final Cut, etc. | AI tool markers |
| Frame rate | Standard (24/30/60fps) | May be unusual |
| Bitrate pattern | Variable (VBR) | Often constant |
Weight: 0.25
Temporal Pattern Analysis
Section titled “Temporal Pattern Analysis”AI-generated videos often have temporal inconsistencies:
- Frame timing: Irregular intervals
- Motion blur: Incorrect blur patterns
- Flickering: Subtle frame-to-frame inconsistencies
Weight: 0.15
Encoding Signature
Section titled “Encoding Signature”Identifies markers from AI video tools:
- Sora
- Runway Gen-2
- Pika Labs
- Stable Video Diffusion
- DeepFake tools
Weight: 0.15
Bitrate Consistency
Section titled “Bitrate Consistency”Real videos have natural bitrate variation. AI videos may have:
- Too-consistent bitrate
- Unusual compression patterns
- Incorrect keyframe intervals
Weight: 0.10
Audio Track Analysis
Section titled “Audio Track Analysis”If audio is present:
- Lip sync consistency
- Audio-video timing
- Audio authenticity (uses audio detection)
Weight: 0.20
Visual Forensics
Section titled “Visual Forensics”When frame analysis is enabled:
- Face consistency across frames
- Background stability
- Edge artifacts
- Reflection/shadow consistency
Weight: 0.15
API Usage
Section titled “API Usage”Upload Video File
Section titled “Upload Video File”curl -X POST http://localhost:8080/verify \ -F "video=@clip.mp4"Response
Section titled “Response”{ "id": "hm_video789", "verdict": "ai", "confidence": 0.78, "content_type": "video", "signals": { "metadata_score": 0.40, "container_analysis": 0.65, "temporal_pattern": 0.72, "encoding_signature": "runway_detected", "bitrate_consistency": 0.55, "audio_analysis": { "present": true, "verdict": "ai", "confidence": 0.80 } }, "processing_time_ms": 350}Current Status
Section titled “Current Status”Video benchmark is pending - requires video file downloads via API keys.
Expected Performance
Section titled “Expected Performance”Based on related benchmarks:
- Target Accuracy: >75%
- Off-the-shelf detectors: 21.3% lower accuracy on Sora-like videos
- Primary Challenges: New diffusion video models, compression
Supported Formats
Section titled “Supported Formats”- MP4 (.mp4)
- WebM (.webm)
- MOV (.mov)
- AVI (.avi)
- MKV (.mkv)
Limitations
Section titled “Limitations”- Full frame analysis requires ffmpeg for extraction
- Re-encoded videos lose original signatures
- Short clips (<3 seconds) have insufficient data
- New AI models may not be in signature database
Best Practices
Section titled “Best Practices”- Original files: Use source video when possible
- Minimum length: At least 3 seconds for reliable detection
- Include audio: Audio track improves detection
- Avoid re-encoding: Each re-encode removes markers
Planned Improvements
Section titled “Planned Improvements”- Full frame extraction with ffmpeg integration
- Face consistency analysis across frames
- Lip sync verification
- Temporal coherence scoring
- Real-time streaming detection
Deepfake-Specific Detection
Section titled “Deepfake-Specific Detection”For face-swap deepfakes, Manuscript looks for:
- Blending boundaries around face edges
- Skin tone inconsistencies
- Eye/teeth artifacts
- Temporal face jittering
- Lighting inconsistencies