The watermarks on Gemini and Nano Banana images

The small four-pointed sparkle in the corner is the only Gemini watermark you can see. A Nano Banana image can carry two more: SynthID in the pixels, and a C2PA Content Credentials record in the file. Almost every “Gemini watermark remover” deals with the first one only, and says little about the other two.

Three marks, in three different places

The rows are independent. Hiding the sparkle leaves SynthID and the credentials in place, stripping the credentials leaves the sparkle and SynthID, and a screenshot throws away the credentials while copying the other two along with the picture.

The sparkle is now a setting

In November 2025, with Nano Banana Pro, Google said it would keep the visible sparkle on images made by free and Google AI Pro users and drop it for Google AI Ultra subscribers and in Google AI Studio. In August 2026 it went further, with a Media Watermark setting in Gemini that turns the visible mark on or off for images, video and music; the first reports of it we found are dated 14 August. Google’s help page for it is plain about the limit: the setting “only controls the visible watermark. It doesn’t affect SynthID watermarks or Content Credentials”. Users in India, South Korea and Vietnam only see the option with an Ultra subscription.

WipeTheAI does not remove visible logos, the sparkle included.

We don’t detect it, paint over it or crop it. For new images the setting is the better route anyway, since it gives you an untouched picture instead of a patched corner. Our view is that most dedicated sparkle removers now solve a problem Google lets you avoid at the source, and their marketing tends to blur the fact that SynthID and the credentials are still there afterwards.

Does your Nano Banana image have Content Credentials?

It depends on the model. Nano Banana Pro (Gemini 3 Pro Image) output has carried C2PA since the week of 20 November 2025, in the Gemini app, Vertex AI and Google Ads. Its February 2026 post on Nano Banana 2 (Gemini 3.1 Flash Image) describes SynthID paired with C2PA Content Credentials. The August 2025 launch post for the original Nano Banana (Gemini 2.5 Flash Image) mentions only SynthID, and we haven’t confirmed whether images made with it before November 2025 carry a manifest.

Look at the file. Our tool reads the file on your device and lists any C2PA manifest with its claim generator and declared actions, and when the file names Gemini, Imagen or Nano Banana it marks SynthID as Likely. It never reports SynthID as detected, for reasons set out in the guide to SynthID checks. For other readers, Adobe’s Inspect site among them, see how to check for Content Credentials.

What each kind of tool does to a Gemini image

• Sparkle removers clean up the corner. The SynthID mark and any C2PA record are untouched, so Gemini’s own check should still call the image Google-made.
• Metadata cleaners such as ExifTool delete the C2PA record and any EXIF or XMP data without changing a pixel. The sparkle stays and so does SynthID. Our tool runs the same clean first, but both of its modes then change the pixels, so what you download from us is never the metadata-only file. How that deletion works per format is in the C2PA removal how-to.
• Screenshots lose the file record too, and keep everything visible or hidden in the pixels.
• Regeneration, our Maximum disruption mode, redraws the picture lightly to disrupt SynthID. It is built to keep what you can see. We haven’t tested what it does to a sparkle, and would expect the mark to survive. It passed Gemini’s provenance check (a classifier judgment, not Google’s production detector) on our test images; results vary per image, and the test notes list what it costs.