Upload a video and DeepSight will analyze it frame-by-frame for temporal inconsistencies, visual artifacts, frequency anomalies, and metadata irregularities commonly found in AI-generated content.
Drop video file here
or click to browse · MP4, WebM, MOV · Max 200MB
——
Initializing analysis…
Extracted Frames
—
—
—
—
Detailed Findings
⚠ Research Prototype — DeepSight uses heuristic frame analysis (temporal consistency, frequency-domain artifacts, edge coherence, metadata inspection). It is not a production forensic tool. Results are probabilistic estimates, not definitive conclusions. Built by Khadija Taki as a cybersecurity research project at Carnegie Mellon University.