Image Forensics - Free Online Photo Analysis Tool
Image Forensics
Inspect metadata and run tamper-detection analysis passes without sending the image anywhere.
Drop an image here
or click to browse from your device
Choose imageCurrent analysis
Metadata
Hover over the analysis output to inspect pixels up close.
Idle.
JPEG Ghost
Re-compress at a chosen quality and reveal regions that respond differently.
Bright, warm contours indicate stronger differences after re-compression. Treat them as clues, not proof.
Frequency Spectrum
View the image’s spatial frequency energy to spot repeating patterns and resampling artifacts.
The center represents broad, low-frequency structure; outer areas represent fine detail and texture.
Block Analysis
Highlights JPEG 8×8 block-boundary discontinuities to reveal an inconsistent block grid.
A uniform grid suggests consistent JPEG compression; faint, shifted, or missing patches of the grid can indicate spliced or resaved regions.
Quantization Table Fingerprinting
Estimates the JPEG quality and encoder signature from the file's embedded quantization tables.
Bit Plane Viewer
Isolates a single bit plane of one channel — the classic way to spot LSB steganography and recompression artifacts.
Plane 0 is the least significant bit; untouched photos usually look like pure noise there.
Per-Channel Histogram
Value distribution of the original image; gaps or comb patterns can indicate prior edits.
Content Credentials (C2PA)
Reads any embedded C2PA manifest's claims — generator, edit history, and named assertions.
AI Upsampling Artifact Check
Scans the frequency spectrum for the periodic checkerboard signature that GAN/diffusion decoders and other upsamplers leave behind.
Highlighted dots mark frequencies far brighter than their neighborhood average. A regular, off-center cluster of them can indicate AI upsampling — but natural repeating textures (fabric, brick, fencing) can also trigger it, so treat this as a clue, not proof.
Embedded Preview vs Pixels
Compares the JPEG preview stored inside the file against the image it is supposed to represent. Editors that rewrite pixels without refreshing the preview leave the original, unedited scene behind in it.
ICC Colour Profile
Reassembles the ICC profile from its APP2 marker segments and reads the header and tag table, which often name the originating device or software.
IPTC & Photoshop Resources
Reads the IPTC IIM datasets and image resource blocks carried in the APP13 Photoshop segment, including its second embedded thumbnail.
Double JPEG Detection
Entropy-decodes the DCT coefficients and fits a generalized Benford model to each AC frequency's first digits. Recompression at a different quality breaks that fit.
GPS Position & Time
Plots the recorded coordinates on a graticule drawn locally in your browser, and cross-checks the GPS timestamp against DateTimeOriginal.
Understanding Image Forensics
What is ELA?
Error Level Analysis re-compresses a JPEG at a known quality level and highlights the difference against the original. Regions that were edited after the last save tend to compress differently than the rest of the image, showing up as brighter patches in the ELA output.
Metadata & GPS
Most cameras and phones embed EXIF metadata directly in a photo: camera make and model, capture settings, and sometimes precise GPS coordinates. This tool extracts and displays that metadata, including any embedded thumbnail or XMP data, so you can see exactly what a file reveals about its origin.
Clone Detection
Copy-move forgeries duplicate a region of a photo elsewhere in the same frame, for example to hide an object or extend a background. Clone detection scans overlapping blocks of the image for near-identical matches at a different offset, flagging clusters that suggest a duplicated region.
100% Private, In-Browser
Every analysis pass runs locally on your device using the Canvas API. Your photo is never uploaded to a server, so sensitive images stay on your machine from start to finish.
Indicators, Not Proof
Every technique on this page — ELA, noise analysis, PCA projections, clone detection, and the rest — surfaces statistical anomalies that are consistent with editing. None of them can, on their own, prove that an image has been tampered with. Re-saving a photo, resizing it, converting formats, or normal in-camera processing can all produce similar-looking artifacts even when nothing was maliciously altered. Conversely, a careful edit made with tools aware of these techniques can sometimes leave few visible traces. Treat every result here as a lead to investigate further — cross-reference multiple passes, examine the metadata, and where the stakes are high (legal, journalistic, or safety-critical contexts) corroborate findings with a trained forensic examiner rather than relying on any single automated signal.
Key Takeaways
- Analysis passes: Metadata, string extraction, magnifier, level sweep, luminance gradient, ELA, noise, PCA, clone detection, JPEG Ghost, frequency-domain analysis, block-grid analysis, quantization table fingerprinting, bit-plane viewing, per-channel histograms, Content Credentials (C2PA), AI-upsampling artifact detection, embedded-preview comparison, ICC profile and IPTC/Photoshop resource parsing, double-JPEG detection and the GPS plot are all available now.
- Results are indicators, not verdicts: anomalies flagged by any single pass warrant further scrutiny, not an automatic conclusion of tampering.
- Nothing leaves your device: every pixel operation runs client-side in your browser; your photo is never uploaded anywhere.