What is Error Level Analysis?

ELA re-saves your image as a JPEG at a known quality and highlights where the compression error differs from the rest of the picture. Areas that were pasted, cloned, retouched or overlaid with text have usually been compressed a different number of times, so they light up brighter than their surroundings.

Does a bright area prove the photo was faked?

No. ELA is an indicator, not proof. Sharp edges, high-contrast detail, text and freshly rendered graphics naturally show higher error levels. Use ELA to decide where to look closer, then confirm with metadata, reverse image search and the original source.

Which quality setting should I use?

Start at 90–95% with 15–25x amplification. If everything looks uniformly bright, lower the amplification; if the whole image looks black, raise it. Images that have been saved many times at low quality often produce a flat, uninformative ELA map.

What's the difference between SHA-256 and a perceptual hash?

SHA-256 is a cryptographic hash of the exact file bytes — change one pixel or one metadata tag and it changes completely. Perceptual hashes (pHash, dHash, aHash) describe what the image looks like, so a resized, re-saved or lightly edited copy keeps almost the same hash.

How do I read the pHash distance?

It's the Hamming distance between two 64-bit hashes. 0–4 bits means practically the same image, 5–10 usually means an edited or re-encoded copy, 11–18 means possibly related, and anything above that generally means two different pictures.

Can I use this for PNG or screenshots?

Yes, any format your browser can decode works — JPG, PNG, WebP, BMP and GIF. Keep in mind that ELA is designed around JPEG compression, so lossless PNG screenshots give less meaningful ELA maps while the hashes stay perfectly useful.

Check if a photo has been edited

Spot edited or spliced photos with Error Level Analysis, and compare images with SHA-256 and perceptual hashes (pHash, dHash, aHash) — 100% in your browser.

Drop an image, or click to browse

JPG, PNG, WebP, BMP, GIF · up to 10.00 MB on free

What people use it for

Check whether a photo sent to you as evidence has had an object pasted into it.
Look for retouched or added text in a screenshot of a receipt, invoice, or bank transfer.
Verify a press or marketplace image before republishing it.
Find near-duplicate images in a library by comparing perceptual hashes.
Prove two files are byte-for-byte identical with a SHA-256 checksum.
Confirm a delivered asset is really the same shot as the approved version, only re-exported.

This free image forensics tool runs Error Level Analysis (ELA) and computes image hashes entirely in your browser. If you've ever wanted to check if a photo is edited, detect a photoshopped image, or verify whether two pictures are the same, this does the job without installing FotoForensics-style desktop suites, Python and OpenCV, or uploading private material to somebody else's server.

The ELA map recompresses your photo at a quality you choose and amplifies the difference, so regions that were pasted, cloned, retouched, or had text added often glow brighter than the untouched background. Alongside it you get a SHA-256 checksum for exact file matching and three perceptual hashes — pHash, dHash and aHash — for finding near-duplicate images even after resizing, cropping or re-saving. Drop in a second image and you'll see the Hamming distance plus a plain-language verdict.

Journalists, moderators, insurance and claims teams, marketplace sellers, photographers and researchers use this kind of photo manipulation detector as a fast first pass. Treat the output as a signal rather than a verdict: ELA highlights where to investigate, and hashes tell you whether two files are the same image, a re-encode, or something entirely unrelated. Everything stays on your device, so confidential images never leave your machine.

How it works

1
Drop your image
Drag in a JPG, PNG, WebP, BMP or GIF, or click to browse. The file stays on your device.
2
Tune quality and amplification
Pick the recompression quality (90–95% is a good default) and how strongly to amplify the difference map.
3
Read the results
Inspect the ELA heatmap, copy the SHA-256 and perceptual hashes, and add a second image to compare similarity.

Frequently asked questions

No. Error Level Analysis, hashing and comparison all run locally in your browser using canvas and the Web Crypto API. Nothing is sent to a server, which matters when you're analysing sensitive or unpublished material.

Want this converter on your website?

Pro members can embed any Minitooler converter on their own site — styled to match, with no branding beyond a small credit link. One iframe, unlimited use, still 100% in the browser.

Unlock embeds — $5/mo

Accent color

Theme

Background

Text color

Width

Height

Embeds are a Pro feature

Upgrade to unlock a customizable iframe snippet you can paste into any website — colors, theme, and size all tuned to match your brand.

Go Pro — $5/mo