EditSpot — find hints of editing in an image

▶ Use it now: https://ronsuthlor3-lab.github.io/editspot/

ภาษาไทย

Shows where an image may have been edited. Runs entirely in the browser. Images are never uploaded. Free, no sign-up.

⚠️ Hints, not proof

Every analysis here only shows that "this area behaves differently from the rest of the image". Edits cause that, but so do text, sharp edges, different textures and resizing. Never judge or accuse anyone on this alone.

Views

View	What it shows	Limits
ELA	Re-saves as JPEG and shows how much each area changed	Best on JPEGs not re-saved many times. Edges/text are naturally bright.
Noise	Fine-grain noise per 8×8 block; pasted or blurred areas often differ	Flat areas (sky) are naturally low
Copy-move	Areas that appear twice in one image (clone stamp, duplicated objects)	Misses copies of flat, featureless areas; natural repeats (bricks, tiles) may be flagged
File info	Camera, capture time, software, estimated JPEG quality, editor traces	Social media usually strips this
Inkseal	If sealed with Inkseal: authentic/modified and the hidden creator ID	—

Copy-move test results (synthetic test images saved as JPEG q70–85): every planted copy was found, including a 60×50 px patch, with zero false alarms on 16 unedited images. Real photos are harder than test images, so treat this as indicative, not a guarantee.

How it works

• src/analysis.js — ELA difference, high-pass noise map, and copy-move detection (downscaled grey image, 8×8 blocks described by mean + 4×4 cell pattern, lexicographic sort with tolerant neighbour comparison, shift-vector voting with sub-pixel bin merging, flat blocks skipped).
• src/jpeginfo.js — EXIF (make, model, software, dates) and quality estimation against scaled IJG tables.
• src/inkseal/ — copied from Inkseal to read seals and hidden IDs.

Run locally with any static server. Tests: npm test (Node 20+).

License

MIT. Part of an artist toolkit with Inkseal, SlipCheck and ArtQueue.