What an invisible watermark is for
You send concept art to a publisher, share a sprite sheet with testers, or upload to a marketplace, and later it shows up somewhere it should not. A visible watermark ruins the art and is easy to crop out. An invisible mark sits inside the pixels: the image looks exactly the same, but it carries a tag only you can read. Give each person a copy with their own tag, and if one leaks you can read the tag and know whose it was.
How it works
Every pixel has red, green and blue values from 0 to 255. Changing the lowest bit of a value shifts it by one step, which the eye cannot see. The tool writes your mark into those lowest bits across the image, after a small header so it can find and verify the mark when reading. Transparent and semi-transparent pixels are skipped so edges stay clean. Because PNG is lossless, saving and copying the file keeps the bits intact.
Reading a leaked copy
Switch to "Read a mark" and drop a file you marked earlier. If the hidden header is found, the tool shows the mark, including the recipient tag if you used per-recipient copies. This is how you match a leaked image back to the person you gave it to.
What breaks the mark
The mark lives in exact pixel values, so anything that rewrites them removes it: saving as JPG or WebP, resizing, filters, or screenshotting. Someone who knows this technique can also strip the low bits on purpose. So treat it as proof of ownership and a tracer for ordinary sharing, not as something that survives a determined attacker. For art you must keep unreleased, the safest step is not to hand out the full-resolution file at all.
Is anything uploaded?
No. Images are marked and read in your browser with a canvas, and nothing is sent anywhere.