Google is testing a way for users to generate images with Gemini’s Nano Banana model without the small visible watermark that currently appears in the lower‑right corner of each picture.
Watermark setting spotted in latest app version
In the most recent build of the Google app, version 17.43.11, an option labeled “Watermark settings” appears alongside other Gemini controls. The setting is listed in the same menu where users switch Gemini accounts, indicating it could be added to the interface if the feature rolls out. The presence of the option has been corroborated by a teardown of the app’s code, which also references “Robin,” the internal codename for Gemini, suggesting the watermark feature is part of a broader push to refine the model’s utilities.
The teardown indicates the new setting would likely target the visible Gemini icon that overlays generated images. That icon can obscure detail in some pictures, prompting frequent requests from users who rely on the Nano Banana model for creative work.
Background on Google’s watermark approach
The company has used two types of watermarks for AI‑generated images. The first is an invisible “SynthID” embedded at the pixel level, which helps other tools verify whether an image was created by AI. The second is a tiny visual overlay that identifies the source as Gemini. While the invisible mark remains in place, the visible logo has drawn criticism for its impact on image composition.
Related: New Chip Hurts Galaxy Watch Battery Life
A recent user survey led by Josh Woodward, the vice president for Gemini, AI Studio, and Google Labs, found the removal of the visible watermark to be the most‑requested change among the top ten suggestions. Woodward said the team would consider the idea, and the appearance of the “Watermark settings” option suggests the suggestion is being acted upon.
Google’s AI Studio already produces images without a visible stamp, aligning Gemini with competitors such as MidJourney and Stable Diffusion, which also rely on hidden watermarks rather than overt logos. The shift could streamline workflows for creators who currently must manually crop or edit out the Gemini icon before using the images elsewhere.
It is not yet clear whether the new setting will apply only to newly generated pictures or if it will enable removal of the logo from existing images. The code review did not reveal any mechanism for retroactive editing, leaving the scope of the feature uncertain.
The company will continue to embed C2PA metadata and the SynthID invisible watermark in all Gemini outputs. Those standards allow downstream tools to verify the provenance of images without relying on a visible cue. The visible logo’s removal would therefore not compromise the ability to track AI‑generated content.
Related: TSMC to generate its own power under new Taiwan law
While the change appears to be in a testing phase, it reflects a broader industry trend toward minimizing visual watermarks while preserving traceability through cryptographic means. Users who frequently employ Nano Banana for marketing graphics, social media posts, or design mockups stand to benefit from a cleaner output that requires less post‑processing.
For those unfamiliar with the technical side, the move illustrates how AI providers balance user convenience with the need for accountability. Invisible watermarks like SynthID are designed to survive typical image edits, ensuring that the origin of a picture can still be verified even if the visible logo is removed.
The company has not announced a timeline for a public release, and the feature may still be in internal testing. Changes typically roll out gradually, often starting with a subset of users before expanding to the broader audience.
Stakeholders will watch the rollout closely, especially as other platforms continue to refine their watermark strategies. Any shift in how the company handles visible branding could influence expectations for AI‑generated media across the industry.