Look for signs a photo was edited TAMPER
Someone sends you a photograph and the question is whether it was edited. TAMPER runs the standard checks and gives you places to look, not an answer.
How it works
What the checks are, and what each one is worth
These are the standard first-pass checks. Every one of them produces places to look, and none of them produces a verdict, which is why this tool does not print one.
- Error level analysis. The image is re-saved at a known JPEG quality and compared with itself. Regions that have been through a different number of compression cycles can stand out. It is the most widely misused technique in this area: a bright region is not evidence of editing, and a uniform image is not evidence of authenticity.
- Cloned regions. Blocks of the image that are unusually similar to other blocks, which is what a clone-stamp or a copy-paste patch leaves behind. Repeating texture (sky, grass, brickwork, a crowd) sets this off constantly and honestly.
- Metadata. The camera make and model, the software field, and the timestamps. An image that says it came from a phone but carries an editor in its software field has been through one. This is the most reliable of the three and the easiest to remove.
When these checks say nothing at all
Most photographs you are handed have been through a process that destroys the signal before you see them.
- Anything from a social platform. Facebook, Instagram, WhatsApp, X and the rest re-encode every upload and strip the metadata. Error level analysis on a WhatsApp image is analysis of WhatsApp's encoder.
- Anything screenshotted or re-saved. One more save at a different quality flattens the history the technique reads.
- A PNG. The technique is about JPEG compression. A PNG has none, so there is nothing to compare.
- Anything generated rather than photographed. These checks look for a photograph that was altered. They are not a detector for an image that was never a photograph.
What to do with the result
Treat it as a list of places to put your attention, then look at the picture. In practice the things that settle a question are ordinary: a shadow falling the wrong way, a reflection that does not match, a background detail that dates the scene, an edge that is too clean. The checks tell you where to start; your eyes finish.
If it matters, get the original. A file straight off the camera or the phone, with its metadata intact, is worth more than any analysis of a copy that has been through three apps, and if somebody cannot produce the original, that is itself the most useful thing you have learned. EXIF reads what is left in a file and ATTEST hashes it so what you were given can be shown later to be what you analysed.
The guides: How to tell if a photo has been edited and Was this audio recording edited?
Questions people ask about TAMPER
How can I tell if a photo has been edited?
There is no single test. The signals worth looking at are differences in compression error across the image, which show where an area has been through a different number of saves; duplicated blocks, which appear when part of the picture was copied to cover something up; noise that does not match between regions; and metadata naming an editor, or a modified date later than the capture.
Is error level analysis reliable?
Not on its own, and it is badly over-read. Resaving, resizing, screenshotting and passing an image through a messaging app all change compression across the whole picture and produce patterns that look like edits. Treat any result as a reason to look harder at a specific area, never as a conclusion.
Can these checks detect AI-generated images?
Not reliably. They are built to find local inconsistencies within a photograph, and a fully generated image has no original to be inconsistent with. Some generators write metadata or provenance credentials, which is a far stronger signal when present, and whose absence proves nothing at all.
What does TAMPER actually check?
Four things: how the picture responds to being recompressed (the error level map), whether any part of it is a copy of another part, whether the JPEG block structure is intact, and what the file says about the software that made it. Each result is a prompt to look harder at a specific spot. None of it is proof, and none of it belongs in an argument as proof. The image is read in this tab.
Does a clean result mean the photo is untouched?
No. Saving the picture once more, taking a screenshot of it, or posting it through any social network erases the error level signal completely. A clean map says only that nothing stood out in the copy you have. Equally, a bright patch is not evidence of forgery: sharp edges, text, logos, a pasted region, a resized region and a dozen innocent edits all show up bright.
Why does it report copied blocks in a photo nobody edited?
Because some things genuinely repeat. Brickwork, foliage, windows and any repeating texture look like cloned blocks to the detector, and it will say so. Treat a report of copied blocks as a place to look, not as a finding.
Doing this often? OBSCURA Case keeps photographs with the rest of the evidence: searched, tagged, with an audit trail and a chronology. Everything is US$5 a month or US$48 a year.