Detection

Analyse a video.

Upload a clip with a visible face. Analysis runs locally and nothing is retained after the result is returned.

Upload

Analysis runs on your own machine. The upload is deleted from disk as soon as the result is returned.

Model

Backbone
CLIP ViT-L/14
Tuning
LayerNorm · 103K params
Trained on
FaceForensics++ c23
In-dataset
0.988 AUC
Cross-dataset
0.942 AUC

Result

No signal

Awaiting input

Select a video on the left and run the analysis to see the verdict, the evidence and the per-frame breakdown.