Platform
A full forensic stack, not a single classifier
FlagFake runs dozens of independent detectors across pixels, frequency domains, metadata and document structure, then reconciles them into one explainable verdict.
One dashboard, every detector
Every check runs the same forensic stack
Confidence score, manipulation-risk breakdown and an ELA heatmap overlay, all in one report you can hand to a reviewer.
Image
Synthetic image forensics
- EXIF metadata & compression analysis
- GAN and diffusion model detection
- Copy-move and clone detection
- Lighting, shadow and reflection consistency
- Error Level Analysis and noise fingerprints
Video
Deepfake video analysis
- Frame-level inconsistency detection
- Face swap and lip-sync mismatch
- Temporal artifact and motion analysis
- Audio-video desynchronisation
- Re-encode and compression signatures
Audio
Voice authenticity
- Voice clone and TTS detection
- Spectrogram anomaly analysis
- Background noise fingerprinting
- Voice biometric comparison
- Splice and edit point detection
Documents
Document tamper detection
- OCR extraction and field validation
- Font and layout consistency checks
- PDF object and metadata forensics
- Digital signature verification
- Inserted image and copy-paste detection
Every result is explainable
Reports include an authenticity verdict, confidence score, risk level, highlighted suspicious regions, a heatmap and a plain-language technical explanation you can hand to a regulator, an auditor or a court.
Output
Confidence scoring
Separate AI-generation and manipulation probabilities, plus a combined confidence score with the evidence behind it.
Output
Heatmap evidence
Suspicious regions are localised and rendered over the original file so a reviewer can see exactly what triggered the verdict.
Output
Signed PDF report
A downloadable forensic report and verification certificate with hashes, timestamps and chain-of-custody notes.