Google is rolling out a new Media Watermark setting in Gemini and Google Flow, allowing users to turn off visible watermarks on AI-generated images, videos, and music. While the visible sparkles disappear, invisible SynthID watermarks and C2PA metadata remain embedded in every file to preserve content provenance.
Generative artificial intelligence tools have long wrestled with a fundamental tension: how to give creators clean, professional-grade media while assuring the public that synthetic creations do not masquerade as reality. Google is now shifting its strategy to resolve that friction. Rather than stamping every creation with an unremovable brand mark, the company is putting visible watermarks behind a toggle switch for users, though important geographic exceptions apply.
How the New Media Watermark Setting Works in Gemini
The updated option carries the name Media Watermark and is located near the bottom of the Settings menu under the left navigation bar. Users can hover over the setting and choose between On and Off states. When left active, the software continues to stamp a visible mark on every asset. Flipping the toggle off yields a clean output file across multiple generative models.
The control covers three distinct models handling different media formats. Images created with Nano Banana, videos generated by Omni, and music produced via Lyria all fall under the new toggle. In addition to the main Gemini application, the setting extends to Google’s Flow filmmaking and video editing tool, where users can toggle off visible marks through their profile menu.
Invisible Provenance Layers Remain Intact Behind the Scenes
Disabling the visible sparkle icon does not mean the file loses all traces of its artificial origin. Google maintains that invisible tracking features continue to function whether the visible toggle is switched on or off. Every downloaded asset retains an invisible SynthID watermark alongside C2PA Content Credentials metadata embedded directly in the background.
Photo: Inshorts
“We’re striking a balance here between creative control and safety: while the visible watermarks are now optional, invisible SynthID watermarks and C2PA metadata are still being used for transparency. So you can still use Gemini or Search to see if an image was AI-generated.”
Josh Woodward, VP of Google Labs, the Gemini app, and AI Studio, via Search Engine Journal
This dual-layer approach allows users to upload supported files directly back into Gemini or search tools to verify original provenance. Verification features such as About this image, AI Mode, and Lens continue to recognize the embedded data layer. Furthermore, Google open-sourced Credentio, a C++ library designed to let developers validate C2PA Content Credentials locally on their own servers without uploading media files to cloud endpoints.
Where the Toggle Is Restricted by Regional Laws
The option to strip away visible watermarks is not universally available. Google confirmed that the setting will be withheld in countries where local regulations mandate visible labeling on synthetic media. Documentation for Google Flow indicates that users creating videos in India, South Korea, and Vietnam will continue to see mandatory visible watermarks, pointing to a similar geographical enforcement pattern for the new Gemini toggle.
For standard users, the update responds directly to consumer feedback gathered by Google executives, answering one of the top user complaints regarding practical utility in professional presentations and marketing materials.
Industry Shifts Toward Invisible Metadata Standards
By removing visible hurdles for everyday creators, Google is betting that invisible cryptography can successfully maintain public accountability. Whether casual viewers will adopt the extra verification steps required to check unlabelled media remains the central unanswered question as the feature rolls out across Gemini, Flow, and eventually Google Search.
