On a week when AI watermarking is a hot topic, Google has just announced that Gemini will let you drop the visible watermarks from created images and videos.
Of course, “visible” is the key word here, with Google exec Josh Woodward explaining in a social post that invisible SynthID watermarks and metadata will remain embedded in created Gemini images.
“We’re striking a balance here between creative control and safety,” Woodard wrote on X, adding that the option to drop the visible watermark is available for the Nano Banana image model, the Omni model for videos, and Lyria for generated music.
You can find the option in your Gemini settings; just click Media watermark, then select either On or Off. I tried it myself (above), and as promised, the resulting Gemini image landed without a Gemini logo watermark.
A new Gemini toggle now lets you determine whether this watermark appears on Gemini-created images.
Ben Patterson (Gemini)/Foundry
Previously, all Gemini-generated images and videos came with a telltale Gemini icon in the bottom-right corner, which served as a visual cue that the content was generated by AI. The ability to remove visible watermarks from Gemini-created images popped up in a top-10 list of requested Gemini features last month.
The new Gemini watermark toggle won’t appear in countries where visible AI watermarks are required, Woodard wrote.
It’s worth noting that ChatGPT’s image creator doesn’t stamp visual watermarks on generated images or videos at all, although it does embed such content with invisible watermarks. The Claude desktop app doesn’t offer a native image generator.
The timing of Google’s announcement about its visual watermark toggle is a little awkward, given the hubbub this week over Anthropic’s news that Claude models will soon embed all generated text with invisible watermarks.
Anthropic’s move comes in response to the EU AI Act, which went into effect earlier this month. The new EU rules mandate that AI providers roll out methods for marking AI-generated content.
There’s a growing consensus that AI-created or edited images, videos, and text should bear at least some type of detectable watermark, allowing those who view or read them to know the “provenance” of the content.
