How to Remove the Gemini “Nano Banana” Watermark

“Nano Banana” is the nickname for Google’s Gemini image model — the one behind the AI images you create in the Gemini app. Like all Gemini/Imagen output, its images are marked in three separate ways, and you have to deal with each one differently.

The three marks on a Nano Banana image

• The visible sparkle. A small four-point star in the bottom-right corner — painted onto the pixels, so it lifts off cleanly. See removing the Gemini watermark.
• Hidden metadata. C2PA Content Credentials plus a “Made with Google AI” provenance tag inside the file — strip it like any C2PA metadata.
• Invisible SynthID. A watermark baked into the pixels themselves. It’s the stubborn one — read how SynthID works.

How to remove each

Start by checking the image — Unmarkr reads the metadata right in your browser and tells you what’s there. Then the visible sparkle and the metadata come off in one pass. The invisible SynthID watermark is the only part that needs heavier work: the image is regenerated on a GPU so the embedded signal no longer survives, giving you a faithful look-alike rather than a byte-identical edit.

Why “just crop the sparkle” isn’t enough

Cropping removes the visible star but leaves both the metadata and the invisible SynthID watermark intact — so the image can still be flagged as AI-made. That surprises a lot of people; here’s why a cleaned image is still detected as AI.