Google's setting that lets you hide the visible corner watermark on media the Gemini app generates — Nano Banana images, Omni video, and Lyria music — while an invisible SynthID watermark and C2PA provenance metadata stay embedded in every file.
Last verified · 2026-08-14 · by Moe Ameen
Gemini Visible Watermark Controls is the setting Google added to the Gemini app on August 14, 2026 that lets you decide whether media you generate shows a visible watermark — the small mark Google stamps in the corner of AI-made images, video, and music. It lives in the app's settings as a Media watermark section with an on/off toggle, and Google described it as balancing creative control with safety. It rolled out over the following days across the Gemini app and Google's Flow video editor, with support in Search planned later.
The control spans Google's generative models by media type: Nano Banana for images, Omni for video, and Lyria for music. Turning the toggle off removes only the visible corner mark. It does not touch the invisible layer: Google keeps an imperceptible SynthID watermark and C2PA provenance metadata embedded in every generated file whether the visible mark is shown or hidden. That means a watermark-free Gemini asset is still machine-detectable as AI-generated — the toggle changes what the eye sees, not what a provenance checker reads.
Access is not universal, and the exceptions are worth knowing before you rely on it. Per reporting on the rollout, the option does not appear in countries whose laws mandate visible AI labeling; in India, South Korea, and Vietnam it is limited to paid Google AI Ultra subscribers, with everyone else in those markets keeping the visible mark automatically; and Google Workspace (work or school) accounts do not get the setting at all. Google has said removing Nano Banana's visible watermark was one of the most-requested changes after the Gemini team asked users about their biggest frustrations.
The honest framing for a creator: this is a labeling control, not a production feature. It gives you a cleaner plate to work from — useful for a client deliverable or a paid post where a competitor's corner logo would look off — but it does nothing to caption, size, brand, or publish the asset, and it does not remove the AI provenance that platforms and detectors actually read. Confirm the current setting, market list, and plan gating on Google's own help pages before building a workflow on the exact details.
Read this as an input-quality decision, not a finish line. The toggle controls one pixel-level detail — the visible corner mark — on a raw asset that still has to become a caption-ready, brand-styled, scheduled post before anyone sees it. That production job is [Kompozy](/), a full AI content generation and multi-platform publishing engine. A clean Nano Banana image drops in as source media and gets composited into a brand-exact [Carousel Post](/glossary/output-buckets), a quote card, or a movable layer inside a [Persona Frames](/glossary/persona-frames) video — all rendered pixel-exact through [HyperFrames](/glossary/hyperframes) — then reframed to 9:16, 1:1, and 16:9, held to your voice by the [Persona Brief](/glossary/persona-brief), and published across the eight social platforms plus blog and email on [Autopilot](/glossary/autopilot).
Here is the specific, non-obvious part: Kompozy already generates face-locked persona images through Google Gemini for its [Persona Photos](/glossary/output-buckets) and Persona Tweets, and it composites brand-facing graphics through HyperFrames rather than posting a raw model image. So a cleaner Gemini plate is genuinely a better ingredient for that pipeline — but Kompozy's value was never the watermark toggle; it is everything that happens after. And because SynthID and C2PA survive the toggle, the disciplined move is to disclose the AI rather than assume a clean corner hides it. Kompozy's per-post review gate is exactly where you set a consistent disclosure line before a single asset ships.
A setting Google added to the Gemini app on August 14, 2026 that lets you hide the visible corner watermark on media Gemini generates — Nano Banana images, Omni video, and Lyria music. It lives in the app's settings as a Media watermark toggle and applies in the Gemini app and Flow, with Search support planned.
No. The toggle removes only the visible mark. Google keeps an invisible SynthID watermark and C2PA provenance metadata embedded in every generated file, so the content stays machine-detectable as AI-generated even with the corner watermark off.
No. Per reporting on the rollout, it is unavailable in countries whose laws require visible AI labeling, limited to Google AI Ultra subscribers in India, South Korea, and Vietnam, and not offered on Google Workspace work or school accounts, where the visible mark stays mandatory.
Mainly for clean deliverables — a client asset, a thumbnail, or a paid post where a visible AI logo would look off. It gives you a cleaner plate to build on. It does not caption, brand, size, or publish the asset, and it does not remove the AI provenance platforms read, so disclosure still matters.
Use a content engine like Kompozy. It composites the clean asset into a brand-exact carousel, quote card, or video layer via HyperFrames, reframes and captions it per platform, keeps it on-brand with a Persona Brief, and schedules and publishes across the eight social platforms plus blog and email — with a review gate for your AI-disclosure line.