Signal Engine v2.1 · March 2026

Check Images & Videos for Unusual Technical Signals

This tool scans for hidden technical patterns — things the eye cannot see — in any image or video file. Everything runs inside your browser. No upload. No login. Your file never leaves your device.

This tool highlights unusual technical patterns. It does not determine truth.

100% Browser-BasedNo UploadNo LoginPrivacy FirstZero Data Transmission

How to Use This Tool

1
📁 Upload your file

Drop or select any image or video. It opens locally — nothing is sent to a server.

2
📊 Review the signal summary

See which technical layers show unusual patterns. Read the plain-English summary at the top.

3
🔍 Investigate further if needed

Use the OSINT next steps section if something looks unusual. A single tool result is never a verdict.

Drop an image or video here

or click to browse — Images up to 50 MB · Videos up to 200 MB

Who This Tool Is For

Scascan's Media Signal Analyser is designed as a professional forensic workflow aid, not a consumer novelty.

📰
Journalists

Verify viral images and videos before publication. Surface technical anomalies to flag for further investigation.

🔎
OSINT Investigators

Analyse media provenance as part of open-source intelligence workflows alongside geolocation and reverse search.

🎓
Researchers

Study visual disinformation patterns. Understand how synthetic and manipulated media presents technically.

🛡️
Social Media Moderators

Add technical signal analysis as one layer of a content review process for suspected manipulated media.

🏢
Brands & Agencies

Audit scraped or reused media assets for signs of manipulation or misrepresentation.

👤
Informed Citizens

Better understand what to look for in viral media before sharing — not to get a verdict, but to ask better questions.

How This Analysis Works

Seven independent signal layers run entirely inside your browser tab — no file leaves your device. Each layer examines a different forensic dimension. Weights are dynamically adjusted per file type: JPEG images emphasise Error Level Analysis, PNG files emphasise fractal continuity, videos add temporal coherence, and screenshot contexts reduce unreliable metadata layers.

1
Metadata Fingerprint
~10% base weight

Reads the invisible data embedded in the file — which software saved it, GPS location if present, timestamps. Files shared on social media usually have this stripped, so the weight is reduced automatically for those cases.

2
Texture Smoothness Check
~18% base weight

Looks at small regions of the image for areas that are unnaturally smooth or uniform. AI-generated images often look perfectly smooth in ways that real-world photos — with natural noise and imperfections — do not.

3
Frequency Pattern Scan
~8% base weight

Looks for invisible repeating patterns in the image that can be left by older AI generation pipelines. Less reliable on modern tools, but still useful as a supporting signal.

4
Statistical Texture Analysis (ML)
~30% base weight

Uses machine learning to examine the statistical distribution of pixel values across the image. This is the broadest and most reliable signal layer — it works across all file types and survives recompression better than others.

5
Compression History
~10% base weight

Examines the file size and structure for signs of how the image was saved — looking at common AI output dimensions, and how the file has been re-saved.

6
Error Level Map (ELA)
~12% JPEG base weight

Re-saves the image at a lower quality and measures the difference. Real photos show uneven differences between textures and smooth areas. Synthetic images often show a suspiciously even difference map across the entire image.

7
Multi-Scale Continuity
~12% base weight

Checks whether fine detail and broad detail in the image are statistically consistent — as they are in natural photography. AI tools often produce images where fine detail looks rich but broader texture is oddly uniform.

Why This Tool Exists in 2026

In 2026, AI image and video generators have become powerful enough that no tool — ours included — can reliably tell you whether a file was made by AI or a human. The generators have become too good, and they actively minimise the very patterns that detection tools look for.

Rather than make claims we cannot support, we built something honest: a tool that surfaces the technical signals a trained investigator would look at — metadata, compression patterns, texture consistency, and frame variation — and presents them clearly. The interpretation is always yours to make.

Our philosophy

We believe transparency is better than certainty. A tool that shows you the evidence is more useful than a tool that gives you a verdict it cannot back up. Verification always requires multiple methods — this tool is one input, not a conclusion.

Why videos are also analysed across frames

Real video has small natural differences between frames — camera shake, changing light, background motion. AI video generators tend to produce frames that are too similar to each other. The tool measures this consistency and flags when it falls outside the normal range for natural video. This is a signal, not a verdict.

Privacy Architecture

Zero Data Transmission
All seven signal layers run in your browser tab using JavaScript Canvas and TensorFlow.js APIs. No bytes are sent to any server at any point.
No Account Required
No registration, email, or identity verification. Access is immediate and anonymous.
Memory Safety
Video frames are processed and immediately discarded. Blob URLs are revoked after analysis. No data persists in memory after the session.
No Logging
We do not log what files you analyse. Your usage cannot be attributed to you. We have no access to your results.

Frequently Asked Questions

Related Resources

Extend your media verification workflow with these tools and guides.