Showcase investigation
Curated sample case. Educational by design. It uses the same verdict UI as the live product while staying explicit about what a demo does and does not prove.
Suspicious short video
The container and preview frame raise provenance questions, but the verdict stays bounded because this MVP does not run full temporal forensics.
Request the original source file or posting context before you circulate the clip as verified evidence.
This demo shows preview-frame and file-structure review only. It is not a full frame-by-frame deepfake determination.
The top evidence surfaced in under five seconds, with the full record below.
Only a representative extracted frame was available for visual scoring.
The preview frame shows enough synthetic-looking structure to justify caution, but not enough for certainty.
These are calibrated product heuristics derived from the inspected evidence, not proof of authenticity or deception.
The trust score combines scam risk, manipulation risk, provenance, claim consistency, and AI-likelihood when it is available.
Request the original source file or posting context before you circulate the clip as verified evidence.
This demo shows preview-frame and file-structure review only. It is not a full frame-by-frame deepfake determination.
- VIDEO demo input loaded from the Dispel showcase dataset.
- Verdict receipt, score model, and evidence hierarchy rendered through the same analysis UI used for saved app results.
- This demo shows preview-frame and file-structure review only. It is not a full frame-by-frame deepfake determination.
- Demo investigations are educational examples, not live user uploads.
Only a representative extracted frame was available for visual scoring.
The preview frame shows enough synthetic-looking structure to justify caution, but not enough for certainty.
No representative frame was available for auxiliary video review, so frame-level signals may remain unknown.
This MVP does not run full temporal or frame-by-frame video forensics. It combines file structure, metadata, and preview-frame cues only when available.
Only signals that were actually inspected are listed here.
Only a representative extracted frame was available for visual scoring.
The preview frame shows enough synthetic-looking structure to justify caution, but not enough for certainty.