Google will now let you decide whether its AI-generated images, videos, and songs carry the familiar Gemini logo in the corner. The company began rolling out a new Media Watermark toggle on August 14, announced on X by Josh Woodward, Google's Vice President for Gemini. The switch removes the visible watermark from content made in the Gemini app and in Flow, Google's AI video editor, with support for Google Search "coming next."
Only the mark you can see goes away. The invisible SynthID watermark and C2PA metadata stay inside every file, so AI-made content can still be identified.
How the Media Watermark toggle works
Open Settings in the Gemini app or on gemini.google.com, then look for Media Watermark. It is a simple on-or-off switch. Turning it off removes the visible logo from:
- Images made with Nano Banana, Gemini's image model
- Videos made with Omni, its video model
- Songs made with Lyria, its music model
The setting works in Gemini and in Flow for now, and the change is rolling out over a few days from August 14. Neither Woodward's announcement nor early coverage mentions a paid-plan requirement for the toggle itself — it appears as a standard setting in the app.
Until now, everything Gemini generated shipped with a small Gemini icon in the bottom-right corner as a visual cue that the content came from AI. Removing it has been one of the most requested Gemini features, and in PCWorld's hands-on test, writer Ben Patterson flipped the new switch and confirmed his next Gemini image arrived with no logo at all.
The hidden SynthID watermark is not going anywhere
Google is careful to frame this as a visible-only change. Every generated file still carries SynthID, Google's invisible watermark, plus C2PA metadata — "content credentials" that record how a file was made. You can still ask Gemini, or use Google Search, to check whether an image was produced with Google's AI tools, TechCrunch reports.
"We're striking a balance here between creative control and safety," Woodward wrote in his announcement.
The move came a day after Google open-sourced Credentio, a C++ library that lets developers build fast, local checking of C2PA content credentials into their own apps. Google says it is the same code that already powers about 40 of its C2PA-enabled products, which have scaled to tens of billions of generated files.
Where the toggle will not appear
Watermark removal will not be offered in countries where the law requires visible labels on AI-generated content, according to Woodward. Google has not published a list of affected countries.
The timing is notable. The European Union's AI Act, which took effect earlier this month, requires AI providers to mark AI-generated content, and Anthropic said the same week that its Claude models will soon embed invisible watermarks in generated text. Google is moving in a different direction on the visible layer — PCWorld points out that OpenAI's ChatGPT image generator has never stamped a visible watermark on its output at all, relying on invisible marks only.
What this means for Filipino users and creators
The Gemini app is fully available in the Philippines, and image generation with Nano Banana works even on the free tier. Heavier users can pay for Google's locally priced AI plans: Google AI Plus at ₱285 a month, Google AI Pro at ₱1,100 a month, and Google AI Ultra from ₱5,999 a month, with Flow video credits bundled into each tier.
No Philippine rule in force today requires visible watermarks on AI media, so the toggle should be part of the normal rollout here — watch for the Media Watermark option under Settings in the coming days. Google has not announced any Philippines-specific restriction.
For freelancers, online sellers, and small businesses that use Gemini for marketing images, product mock-ups, or short videos, the change means cleaner output without cropping the logo or using third-party removal tools. Cleaner does not mean untraceable, though. Philippine authorities are already watching this space: the National Privacy Commission has warned that AI images of real people can break the law, and every Gemini file still carries SynthID, so misuse remains detectable.
The toggle also lands in a busy month for Google's AI push, arriving a day after Gemini 3.7 Flash launched with major coding gains at half the price of its predecessor.
