Research
New methods make it easier to detect deepfake videos
Researchers present a combination of image analysis and pattern detection that identifies AI-generated video more precisely than earlier methods.

This article is a fictional example created for demonstration. People, events and figures are invented.
The imagined study combines analysis of single frames with analysis of motion over time.

Results (fictional)
Method | Precision |
|---|---|
Frame analysis only | 82% |
Combined method | 93% |
Show technical analysis
The model uses a convolutional per-frame detector and a temporal model that looks for inconsistency between frames.
Sources
- Fiktiv demokilde – ingen ekte artikkel – Deepfake.no (eksempel)