Compare two skin revisions pixel by pixel
Minecraft Skin Diff compares an older and newer 64 × 64 skin PNG at every aligned atlas coordinate. It counts exact RGBA changes, separates added visible pixels, removed visible pixels, and alpha changes, then produces a transparent heatmap. Green marks additions, red marks removals, and yellow marks pixels that remained present but changed color or opacity.
Both files are decoded locally. Their names, pixels, difference report, heatmap, and CSV never go to a server. The page starts with a first-party pair containing controlled changes, so all statistics and downloads can be tested immediately.
How to use the comparison
Upload the older PNG into the first slot and the newer PNG into the second. Both must be exact 64 × 64 PNG files using the same atlas alignment. The previews display each revision and update the report as soon as either input changes.
Download the heatmap when reviewers need a visual overlay. Download the CSV when they need exact x and y coordinates plus before and after RGBA values. The CSV includes one row per changed pixel and can be opened in a spreadsheet, attached to an asset review, or parsed by a build script.
If the two files use different layouts, convert them first. A 64 × 32 legacy texture and a 64 × 64 modern texture are not coordinate-comparable even if they depict the same character. Use the legacy converter before running an aligned diff.
Reading the heatmap colors
Green means the older pixel had alpha zero and the newer pixel became visible. Red means a visible older pixel became fully transparent. Yellow means both versions differ without fitting those two cases: RGB changed, alpha changed while both remained visible, or hidden channel data changed.
The heatmap is diagnostic, not a playable skin. Unchanged pixels are transparent so the marks can be layered over another preview in a desktop editor. Do not import the heatmap as a character texture.
Added and removed counts focus on visibility. The alpha-change count includes additions, removals, and transitions between nonzero alpha values. Therefore categories can overlap; they are not intended to sum to the total changed count.
Exact RGBA comparison versus visual similarity
The engine compares channel bytes, not human perception. Changing red from 120 to 121 counts as a difference. A hidden RGB value beneath alpha zero can also differ even though neither pixel is visible. This strict behavior is useful for regression testing, export verification, and controlled reviews.
It does not measure whether two skins “look the same,” identify copied artwork, or calculate perceptual similarity. It also does not compensate for shifted atlases. If an editor moved a face by one pixel, every displaced coordinate can be reported.
Use the Minecraft skin palette extractor when the question is about overall color frequency rather than coordinate changes. Use the skin compatibility checker when the question is format suitability.
Practical revision workflows
Commission reviews benefit from exact evidence. A client can compare the delivered revision against the approved source and see whether edits were limited to the requested area. Teams can archive the CSV beside source files, while artists use the heatmap to locate unexpected changes at atlas scale.
The tool is also useful after a conversion or cleanup. Compare a skin before and after the background cleaner to confirm only unused coordinates were removed. Compare before and after a transparent pixel repair to verify the intended holes changed.
For recoloring, a diff confirms scope. After using the Minecraft skin color replacer, the report should show changes only where matching colors existed. An unexpectedly large count can reveal excessive tolerance or a wrong layer selection.
Coordinate and alpha details
Coordinates are zero-based, so the upper-left pixel is (0, 0) and the lower-right is (63, 63). RGBA values range from zero to 255. Alpha zero is fully transparent and 255 is fully opaque.
The tool does not infer Classic or Slim because an exact image diff does not need a body model. It compares the entire 4,096-pixel canvas, including unused atlas space. This is deliberate: hidden export residue can matter to a byte-level workflow even if it never appears on a standard model.
If the goal is to ignore unused positions and examine only mapped faces, clean both copies under the same model first or filter the downloaded CSV in a separate process.
What the diff cannot decide
A difference is not automatically a defect. Added outer-layer detail, deliberate transparency, shading, or a corrected seam should appear in the report. The tool cannot decide which revision is better or which change was authorized.
It does not validate account ownership, copyright, moderation, game version, server behavior, or skin-pack metadata. A zero-difference result proves equality between decoded RGBA buffers, not platform acceptance.
Privacy and safe handling
Browser canvas APIs perform decoding and rendering without an upload route, account, database, or gallery. Closing the page discards the selected revisions unless reports were downloaded.
Only compare files you own or are permitted to review. A technical difference report does not grant permission to redistribute either skin. SkinEditor.org is independent and is not associated with or approved by Mojang or Microsoft.
Frequently asked questions
Why must both PNGs be 64 × 64?
Exact coordinate comparison requires matching modern atlas dimensions and alignment.
Does yellow mean a bad pixel?
No. It means an existing coordinate changed color or alpha; the reason may be completely intentional.
Can I use the heatmap as a skin?
No. It is a diagnostic transparent overlay, not a complete character atlas.
Does the tool upload both files?
No. Both are decoded and compared locally in the browser.