Independent signals, one cautious report.
Each family answers a different forensic question. The product preserves disagreement, reports missing coverage, and turns technical output into a next action.
AI models
Whole-image, regional patch, and residual-statistics detectors.
Pixel forensics
Noise, FFT, DCT, ELA, copy-move, seams, entropy, and recapture clues.
Document logic
Localization, rectification, two OCR engines, MRZ, barcode, and geometry.
Face checks
Face roles, presenter-to-document comparison, and passive presentation clues.
Provenance
Offline C2PA validation with explicit trust states, metadata, native container parsing, and file integrity.
Audio & voice
Container validation, waveform continuity, spectrograms, AudioSeal watermark localization, and AASIST speech anti-spoof research evidence.
Optional local watermark research
Checkpoint-gated VideoSeal/PixelSeal-compatible detection runs only with an operator-supplied local model; it is not a SynthID detector.
Forgery localization
PhotoHolmes methods and learned visual heatmaps with conservative interpretation.
No static-image detector can prove authenticity. AI Screener combines independent signals, shows the evidence, and leaves the decision with a human.
Run the free checker