Map related versions

Build an image lineage and version graph

Compare 2–40 related images and build an inspectable graph of exact copies, resizes, recompressions, crops and tone-adjusted derivatives.

Map image versions

Completely free No account or payment Image stays on this device

USE THIS WHENUse this when a folder contains originals, crops, social copies, recompressed exports or other versions whose relationship is unclear.SUPPORTED INPUT2–40 related images

What you will getA clear result before technical detail

01A visual map of plausible parent and child versions

02Per-edge similarity scores and transformation evidence

03An exportable JSON report with hashes, dimensions and explicit uncertainty

Private analysis labFiles stay inside this browser

LINEAGE / 40 FILES

Choose related versions of one image2–40 JPEG, PNG or WebP files · originals, crops, resizes and exports can be mixed

Selection is processed locally

Session diagnosticsNo processing errors

Stored only in this browser tab. Image bytes, filenames and metadata are never included.

No tool error has been recorded in this tab.

No upload or account Originals remain unchangedReview methodology

Useful and careful

What is built into the result.

Important findings come first. Technical fields remain available without taking over the page.

Combines exact SHA-256 with three perceptual hashes

Explains every graph edge instead of returning a black-box match

Never presents a probable root as a proven original

Understand the method

A useful result with its reasoning attached.

This page explains what the lab measures, how to interpret it and where human review remains essential.

A relationship graph instead of a one-number comparison

Ordinary image comparison tools often stop at a similarity percentage. That answer is difficult to interpret when a folder contains an original export, a social-media copy, a crop, a WebP conversion and a version with a caption. The lineage lab evaluates the set as a network and proposes the smallest plausible path connecting related files.

The graph begins with exact SHA-256 matches, then combines average hash, difference hash and a low-frequency perceptual hash with aspect ratio, luminance, pixel dimensions and file size. The components remain visible in the evidence table so the final relationship is not an unexplained score.

Exact byte duplicates
Resize or recompression evidence
Crop or layout derivatives
Tone-adjusted derivatives

How a possible root is selected

A larger, information-rich image is evaluated before smaller derivatives because it is a reasonable parent candidate. That ordering is a heuristic, not a timestamp. A high-resolution re-export may be newer than a small original, and metadata dates can be changed or removed.

For that reason the interface uses the phrase possible root. When more than one image cannot be connected above the reporting threshold, the result shows multiple graph components rather than forcing a single origin story.

No automatic original claim
Multiple roots allowed
Threshold-level isolated files remain visible
Chronology warning included in exports

Useful investigations and responsible limits

The graph can organize publisher asset folders, trace web derivatives, compare edits made for different platforms, and help an archive select the highest-information copy. It can also expose when a supposed duplicate is actually a crop or recompressed version.

The lab does not determine whether the scene is true, who owns the copyright or which person edited the file. Content Credentials, trusted storage records and external publication dates remain separate evidence layers.

Editorial asset review
Archive deduplication
Social copy comparison
Before-and-after documentation

Three simple steps

Know what happens before you start.

Use this when a folder contains originals, crops, social copies, recompressed exports or other versions whose relationship is unclear.

FingerprintCompute an exact digest, three perceptual hashes and inspectable image statistics.
ConnectBuild only threshold-level parent-and-child hypotheses while allowing multiple roots.
ReviewInspect every score and export uncertainty alongside the evidence.

RESULT ORDER

01 · A visual map of plausible parent and child versions02 · Per-edge similarity scores and transformation evidence03 · An exportable JSON report with hashes, dimensions and explicit uncertainty04 · Limits and next step

Clear before you rely on it

Questions this tool should answer upfront.

Short answers keep important privacy, evidence and professional-use limits visible.

01Can this tool find the true original image?

It can rank plausible roots using dimensions, file evidence and similarity, but it cannot prove chronology without trusted external records. The result deliberately says possible root.

02What is the difference between an exact and perceptual match?

An exact match has the same SHA-256 digest and identical bytes. A perceptual match can survive resizing or recompression, but it carries uncertainty and can produce false matches.

03Are image previews or hashes sent to AnalyzeImage?

No. Fingerprints, graph construction and previews remain in the active browser tab. The optional JSON report is created only when you download it.

Method referencesColumbia University Visual Migration Maps research C2PA specification

Continue when useful

Your next step, without starting over.

Move to another page only when its outcome matches what you need.

Compare metadataReview what fields changed between two selected copies.Read the edit-evidence guideUnderstand what file evidence cannot prove.

Important limitations

Visual similarity cannot prove authorship or creation time · Large crops, collages and heavy edits can evade the current hash model · A high score is evidence for review, not proof that two depicted events are identical

Full limitations