Google is giving Gemini users a new way to control how AI-generated media looks when exported: a Media Watermark setting that can turn off the visible mark normally applied to generated content, according to Tom’s Guide. The report says the option covers images, videos and music created with Google’s AI tools. Tom’s Guide identifies the affected outputs as images generated with Nano Banana, videos made with Omni and music from Lyria. The setting is rolling out across Gemini and Google’s Flow filmmaking tool, with Google Search expected to receive the same option later. The practical change is narrow but meaningful for creators. Users making presentation assets, mockups or social posts can generate a cleaner file without cropping out the visible Gemini mark or removing it in separate editing software. Tom’s Guide says the setting applies only to content created after the user changes the option; it does not automatically strip marks from files already downloaded. The move does not make Google-generated media unmarked in a provenance sense. Tom’s Guide reports that Google will continue embedding its invisible SynthID watermark and C2PA Content Credentials metadata in generated files even when the visible watermark is disabled. Those machine-readable signals are intended to preserve information about the file’s origin or AI generation history without altering how the image, video or audio appears to a viewer. That distinction matters because visible and invisible provenance mechanisms serve different uses. A visible mark can warn viewers when realistic media is encountered without context, but it can also be distracting when the output is clearly being used as a creative asset inside a larger project. Google’s approach, as described by Tom’s Guide, makes the visible mark optional while retaining less intrusive signals designed for detection and provenance tools. The rollout may not be visible to every user immediately. Tom’s Guide says Google is bringing the feature to accounts over several days, and that users may need to look in the Gemini account menu or Flow’s visible-watermarking controls. The report also notes a country-specific limit: Google’s Flow support page says visible watermarks are applied automatically in India, South Korea and Vietnam, so users in those markets may not be able to remove them. Who benefits: Gemini and Flow users who want cleaner AI-generated assets benefit most, especially when the visible mark interferes with presentations or mockups. Google also keeps a provenance layer in place through invisible watermarking and metadata, according to the report. Who's exposed: Users who rely on visible marks as the primary disclosure mechanism may see less obvious labeling on Google-generated media. The exposure is partly mitigated by the reported persistence of SynthID and C2PA signals, but the cluster does not provide evidence on how consistently those signals will be checked by viewers or platforms.