Google is rolling out a new setting for Gemini that gives users control over the visible watermark applied to AI-generated images, videos, and music. The update, called Media Watermarks, addresses one of the most common complaints from people who use Gemini for creative work. Until now, the visible watermark was automatically applied to every generated file, and there was no official way to remove it.
The change comes after Google's vice president Josh Woodward asked users to share their biggest Gemini frustrations. Thousands responded, and the request to remove the visible Nano Banana watermark quickly landed in the top ten. Now, with the new toggle, users can choose whether their future creations carry a visible mark.
What is changing in Gemini?
Gemini's Media Watermarks setting is designed to cover multiple creative tools inside the app. It applies to images made with Nano Banana, videos made with Omni, and music made with Lyria. According to Google, the setting will roll out over the next few days. Users who want to adjust it can simply go into Gemini's settings, select Media Watermark, and toggle the visible watermark on or off.
The choice applies to every future image, video, and music track created in Gemini. It is a global setting rather than something users can change per file. That means once you turn off visible watermarks, all future generated content in those formats will not display a visible mark by default.
It is important to note that the toggle only controls the visible watermark. Google has made it clear that invisible SynthID watermarks and C2PA metadata will always remain embedded in the background. This means AI-generated content does not become untraceable when the visible mark is removed. The origin of the content can still be identified using the hidden metadata and watermarking technology.
How to turn off AI watermarks for Nano Banana, Omni, and Lyria in Gemini
The process is straightforward. Open the Gemini app and go to Settings. Look for the Media Watermark option. Tap it and toggle the setting based on your preference. The setting applies across images, videos, and music, so there is no need to configure separate toggles for each content type.
Google says the change will become available gradually over the next few days. If you do not see the option immediately, it may not have reached your account yet. Users with compatible devices and accounts should receive the update as Google expands the rollout.
For those who prefer visible watermarks for their own reasons, keeping the setting on is also an option. Some creators use visible watermarks to protect their work from being passed off as human-created or to maintain transparency when sharing AI output.
What stays hidden underneath
Turning off the visible watermark does not mean the file is anonymous. Google uses SynthID, a deep watermarking technology that embeds a subtle pattern or signal into the pixels, audio frequencies, or metadata of generated content. This signal is designed to survive edits, screenshots, and compression, making it difficult for someone to strip it out without damaging the content.
In addition to SynthID, Google also includes C2PA metadata. C2PA, short for Coalition for Content Provenance and Authenticity, provides a standard way to record the provenance of digital content. This metadata can show when a file was created, what tool was used, and other details about its origin. For journalists, researchers, and platform moderators, this kind of provenance is increasingly important in the fight against misinformation.
So while visible watermarks can be disabled, the invisible infrastructure remains. This approach reflects a broader industry trend. Many AI companies are experimenting with hidden watermarking systems that do not ruin the visual experience but still allow for verification.
Regional and account restrictions
Not every user will get the new watermark toggle. Google has placed some limits on the rollout. In India, South Korea, and Vietnam, the option is only available to paid AI Ultra subscribers. Free users in those countries will continue to see visible watermarks on their generated content, and they cannot turn that off unless they upgrade.
Work and school accounts are excluded from the setting entirely. Users signed in through Google Workspace for education or enterprise will notice no Media Watermark option, meaning visible watermarks remain mandatory for those accounts. Google has not shared a timeline for expanding the toggle to these accounts.
This tiered rollout has raised questions about why Google chose to limit access in specific regions. The company has not explained the reasoning in detail, but it may be related to local regulatory expectations around AI content disclosure. Governments and regulators around the world have been introducing policies that require AI-generated content to be clearly labeled, and visible watermarks are one way to meet those requirements.
Why did Google add the AI watermark toggle to Gemini?
Google's decision is closely tied to user feedback. Josh Woodward, a Google vice president, asked the Gemini community what frustrated them most. The response was overwhelming. Thousands of users submitted their complaints, and the visible Nano Banana watermark was among the most popular requests. Woodward acknowledged the request and said the team would look into it. Now, the official option has arrived.
The update also lands as community efforts to remove watermarks have gained traction. Some users have found ways to crop, prompt-inject, or edit around watermarks, while others have called on Google to make the watermarking system less intrusive. By adding an official toggle, Google may be hoping to reduce the incentive for third-party tools that bypass watermarking altogether.
At the same time, Google is likely trying to balance user convenience with responsible AI practices. Visible watermarks are a simple way to alert viewers that a piece of content is synthetic. Removing them entirely could make it harder for casual viewers to know what they are looking at. The hidden SynthID and C2PA layers provide a technical solution, but they are not as transparent to the average person.
The bigger debate around AI watermarks
Watermarking has become one of the most discussed topics in AI policy and product design. Many companies, including Google, Microsoft, and OpenAI, have developed their own approaches to marking AI output. Some systems embed text-based labels, others use cryptographic metadata, and some rely on statistical patterns that are invisible to human eyes.
SynthID is notable because it works directly within the content rather than being a separate metadata file. That makes it more resilient than simple metadata tags, which can be accidentally deleted or deliberately stripped by users. Google first introduced SynthID in 2023 for its image generator, and later expanded it to video, text, and music. The technology can be applied during generation, allowing for a more robust mark that survives common transformations like resizing, color adjustment, and re-encoding.
C2PA, meanwhile, is an open standard supported by a coalition of tech companies, news organizations, and camera manufacturers. It was created to address the growing need for provenance in digital content. C2PA metadata can include information about the device or software used to create a file, the time and place of creation, and any edits made after the initial creation.
The emerging consensus among experts is that robust provenance requires multiple layers. A visible mark can warn people immediately. An invisible watermark can survive tampering. Metadata can provide a detailed record. Together, these systems make it much harder for someone to pass off AI-generated content as authentic human work.
There are also debates about whether visible watermarks are effective at all. Studies have shown that people often ignore watermarks or fail to notice them, especially on busy social media feeds. Hidden watermarks and metadata may be more reliable, but they require technical tools to verify. For that reason, some regulators still prefer visible labels, particularly when it comes to political advertising or realistic deepfakes.
Google's new toggle gives users more creative freedom, but it also places more responsibility on the platforms and users who share the content. A user who disables the visible watermark is not hidden from forensic analysis, but the audience viewing the content may not know it was made by AI. That creates a trade-off between aesthetics and transparency.
The move also highlights how quickly AI product features are evolving. When Gemini first launched, visible watermarks were treated as a fixed policy. Today, they are a user preference. Tomorrow, they could be regulated by law. Companies are adapting to public feedback while also trying to stay ahead of policymakers who are paying close attention to AI risks.
For now, Gemini users who want to remove the visible watermark can do so in just a few taps. The hidden markers remain, ensuring that the content's origin is still discoverable by those who need to know. The rollout is gradual, and exceptions apply in certain regions and account types, but it marks a notable shift in how Google handles AI content disclosure.
Source: Digital Trends News