Free Image Tampering Detector

What does the tampering detector do?

The freeimgtool tampering detector runs Error Level Analysis (ELA), a classic forensic technique for spotting edits in JPEG images. The idea: a JPEG that has been opened, edited, and re-saved compresses differently in the edited regions than in the original. By re-saving the input at a known quality and measuring the difference, edited regions show up as brighter patches in the result image.

ELA is a heuristic, not a verdict. A bright spot does not prove fakery, and a clean ELA does not prove authenticity. Treat the result as a hint — a starting point for closer inspection.

How to use it

1. Click the upload box and pick a JPEG (up to 5 MB).
2. Drag the amplification slider. Start around 15. Raise it to see fainter regions, lower it if the whole image is white.
3. Click Analyze.
4. Inspect the verdict, max/mean error numbers, and the ELA image.
5. Optionally download the ELA visualization to compare with the source.

How to read the result

• Uniformly dark with sharp edges visible: consistent compression history — usually a clean original.
• One area noticeably brighter than its surroundings: that area may have been pasted, masked, or edited.
• Entire image is bright: the JPEG has been re-saved many times. ELA loses signal here.
• PNG or WebP input: ELA cannot meaningfully separate edits from compression because there is no JPEG quantization baseline.

What common edits look like in ELA

• A pasted object (a person, logo, or product added from another photo) usually shows as a patch whose texture and brightness do not match the area around it, often with a crisp outline where the paste was masked.
• Changed text on a screenshot or document — a price, date, or name typed over the original — tends to glow brighter than the untouched text beside it, because the new glyphs have been through one fewer JPEG save.
• Clone-stamped or healed areas (a removed blemish, wire, or person) look smoother and flatter than the surrounding texture, sometimes with a repeated pattern.
• Whole-image adjustments like brightness, contrast, or a filter change the entire image evenly. ELA cannot see them, because nothing stands out from anything else.

Always compare like with like. Edges, text, and fine texture are naturally brighter in ELA than sky or skin. The question is whether one region is brighter than other regions with similar detail, not whether anything is bright at all.

How the analysis works here

Your image is decoded in the browser, scaled so its longest side is at most 1,800 pixels, then re-saved as a JPEG at quality 90. Each pixel of the re-saved copy is compared with the original, and the difference is multiplied by the amplify value so small changes become visible. The verdict (low, medium, or high) comes from the single largest difference in the image, so it is easily pushed up by sharp edges. Treat the verdict as a prompt to look, and the ELA image as the real result.

A five-step check before you trust a photo

1. Find the original. Ask the sender for the first file they received, not a forwarded copy. Every re-save weakens the evidence.
2. Run ELA here and note any region that is brighter than similar detail around it.
3. Reverse image search with Google Lens or TinEye. An older or differently cropped version online is often the fastest proof of an edit.
4. Check light and geometry. Shadows should fall the same way, reflections should match, and straight lines near a suspicious area should not bend.
5. Check the context. Who first posted it, when, and does the scene match the claimed place and date? Our guide on how to spot edited images covers these signs in more detail.

When ELA cannot help

• Photos sent through WhatsApp, Facebook, or X. These apps resize and re-compress every upload, which flattens the differences ELA relies on. A clean result from such a copy proves very little.
• AI-generated images. An image generated in one pass has one compression history throughout, so ELA sees nothing unusual. It detects edits, not synthetic content.
• Screenshots and heavily resized images, which re-render every pixel and erase earlier compression traces.
• PNG, WebP, and HEIC files, which have no JPEG compression history to compare against.

Why use this tool

Most photo forensic tools sit behind academic logins or paywalled investigator suites. ELA is the single most useful check a non-expert can run quickly. This tool gives you the slider and the visualization in three clicks with no signup.

Who uses it

Journalists vetting a forwarded photo. Investigators spotting edits in a contract scan. Insurance adjusters checking damage evidence. HR teams reviewing scanned documents. Anyone who wants a fast sanity check before trusting an image.

Frequently asked questions