SignalLens, the local visual-forensics tool for explainable image analysis, is now available as a Windows early-access beta.
SignalLens is not a simple ‘AI detector’. Nor does it pretend to be a truth machine.
Instead, it analyses multiple forensic signals and produces an explainable audit trail that helps users to reason about camera-origin evidence, synthetic-generation indicators, editing traces, metadata, and conflicting signals.
Even in this first beta, SignalLens already includes a broad set of analysis capabilities.
It is not distributed as a collection of Python scripts that need to be run manually by specialists. Instead, it offers a complete local GUI experience built with the Qt framework.
Local image analysis can optionally start with OSINT, exploring an unknown image’s possible background through reverse image search. We support metadata analysis through ExifTool. Revealing invisible information about image creation, editing, software traces, camera data, and embedded workflow clues.
Frequency-domain analysis using FFT and patch-similarity analysis can help detect anomalies in suspect textures and repeated structures. We use MediaPipe-based segmentation for subject/background and boundary-region investigation. Helping SignalLens examine localized manipulations such as generative fill, inpainting, or face replacement.
To make these signals inspectable, SignalLens produces visual overlays, masks, heatmaps, JSON outputs, summaries, and a full local audit trail for every analysed image.
Each audit trail documents the evidence behind the conclusion, including a reasoning trace that explains the final interpretation.
SignalLens is still beta software. The executable is unsigned. Windows may show a SmartScreen or Smart App Control warning. You should treat results as forensic reasoning rather than absolute proof.
The download is available here.
Feedback is very welcome, especially from users working in OSINT, journalism, computer vision, image analysis, platform moderation, e-commerce, or synthetic media research.
A short version of this post is published on LinkedIn