Gemini Image Watermark Detector: Logo, C2PA & SynthID
Every image from Google's Gemini (including the Nano Banana and Nano Banana Pro models) can carry up to three watermarks: the visible sparkle logo in the corner, an invisible C2PA Content Credentials manifest in the file header, and Google DeepMind's SynthID signal hidden in the pixels. This Gemini image watermark detector checks the visible logo and looks locally for familiar C2PA marker strings. That marker scan is a hint, not manifest verification; the official content-provenance checker is required for a verified hidden-watermark result.
Upload an image to see exactly what it carries. If it's watermarked, you can clean it in one click — the detector hands straight off to the remover. If a platform labelled your post and you are working out why, this is the place to start — the answer is almost always one of these layers.
What the Gemini Detector Actually Checks
The logo is detected locally; C2PA gets a marker-string hint; hidden credentials require an official verifier.
The visible scan uses the same engine that removes the Gemini logo: it looks for the sparkle mark at its known position and reports a match with confidence. The hidden scan is deliberately narrower: it searches the file bytes for strings commonly found around C2PA and JUMBF data. It does not parse a manifest, verify its signature or prove who issued it.
If no familiar marker is found, that doesn't mean the image is clean: a local string scan can miss credentials, and Gemini can also apply SynthID in the pixels. Use the official checker before relying on either a positive or negative hidden-watermark result.
Visible Logo vs C2PA vs SynthID
Three independent layers, three different removal methods — an image can carry any combination.
These are three independent layers. The visible sparkle is pixels you can see and is removed with lossless reverse alpha blending. C2PA is metadata in the header — invisible to you, read instantly by machines, and stripped by re-writing the file. SynthID is invisible and lives in the pixels themselves, so it survives a metadata wipe and has to be dealt with in the pixels. A Gemini image can carry any combination of the three, which is why a single "is it watermarked?" answer needs all three checks.
From Detection to a Clean Image
Detection routes straight to the tool that clears whichever layer was found.
Detection is only half the job. When the detector finds the visible logo, the "Remove the logo" button takes you to the Gemini image remover, which clears it with no quality loss. When C2PA or SynthID is the issue, you're routed to hidden-watermark removal, which processes the picture itself to clear SynthID and drops the C2PA manifest in the same pass. Either way the original image quality is preserved.