Stop guessing if an image is real. Start understanding it.

SignalLens analyzes images using multiple forensic signals and produces a clear, explainable reasoning trace — all locally.

SignalLens
Subscribe for the Upcoming Release
SignalLens

SignalLens

SignalLens - Local Visual Forensics

We’ve lost the ability to trust what we see:

AI images are no longer obviously fake. Real images are edited, filtered, and enhanced.

Tools respond with a single number: AI-generated: 87% That’s not evidence. That’s a guess.

A different approach

SignalLens doesn’t guess.
It analyzes.
  • Extracts multiple forensic signals
  • Interprets them using structured reasoning
  • Produces clear, human-readable explanations

SignalLens Architecture

SignalLens doesn’t just detect – 
it explains

Signal Extraction:

FFT, patch similarity, metadata, boundary analysis

Structured Reasoning:

Signals are interpreted, not averaged

Explanation:

Clear forensic conclusion instead of a score

SignalLens is used for:

• Investigating suspicious images
• Verifying online content
• Understanding AI vs real ambiguity
• Exploring how images are constructed

You don’t just get a score. You get the evidence behind the conclusion.

SignalLens does not only show a verdict: Full Forensic Evidence Package

For every analyzed image, it creates a dedicated local output folder inside: .\analyzed\

Each image gets its own subfolder, containing the full forensic evidence trail: visual overlays, heatmaps, masks, FFT outputs, metadata JSON, patch similarity results, logs, summaries, and the complete reasoning trace.

The output includes:

  • Dedicated Folders: Logs, prepared images, and summaries.
  • Generated Artifacts:
    • ELA overlays and FFT spectra
    • Patch heatmaps and boundary masks
    • Subject and region masks
    • JSON signal files and HTML summaries
    • Complete reasoning traces

signallens
SignalLens © 2026 Privacy-first • Runs locally