How to remove the Gemini watermark (2026)
Gemini marks a generated image twice over. A small sparkle glyph goes into the corner of free-tier output, and an invisible SynthID signature goes into the pixels of every generation on every plan. The visible glyph is easily dealt with: paid Google AI tiers never apply it, and on an image you already hold, Dewatermark.ai detects and clears it free, twice a day in HD.
SynthID cannot be removed and is not worth chasing. It is invisible, it survives editing by design, and it exists so platforms can identify AI output. Everything below concerns the visible glyph, on images generated from your own account.
What Gemini adds
A small sparkle glyph in the corner, over a SynthID layer, on every generated image
- Free tier
- A paid tier clears the glyph; the embedded signature is there for good
- Best remover
- Dewatermark.ai, 8.2/10
Three ways to lose the sparkle glyph
The visible glyph comes off easily. Generating without it is still the better outcome, and costs nothing extra if you already subscribe.
Paid tiers generate images without the visible glyph, at full quality, on everything you make afterwards. The clean fix for anyone using Gemini regularly.
Check Google AI plans (opens in a new tab)Generative Fill rebuilds detail beneath the mark better than one-tap tools, which is what you want when the glyph sits over your subject rather than empty space.
Try Photoshop (opens in a new tab)Finds the corner mark unaided, two free HD exports a day, no brush and no signup ritual. Sufficient for the great majority of Gemini images.
Try Dewatermark.ai (opens in a new tab)Clearing a Gemini glyph with Dewatermark.ai, step by step
The free route, and against a corner glyph it is genuinely a one-minute job. When the mark crosses detail you actually care about, take the Photoshop route instead.
-
Upload the generation untouched
Drop the Gemini image into Dewatermark.ai exactly as it came out of the generator. Editing or compressing it beforehand only discards information the fill could otherwise have used to rebuild the corner.
-
Let detection find the glyph
The model locates solid corner marks like Gemini’s on its own, and our test image was boxed within a couple of seconds. Reach for the manual brush only when the mark overlaps something visually busy and the detector visibly hesitates.
-
Inspect the corner before downloading
Compare the filled corner against the surrounding area at full zoom. AI-generated backgrounds tend to be smooth, which makes fills easy, and any mismatch you can see at 100 % will be equally visible to everybody else.
-
Export, and remember what stays
Download in HD, free twice a day. If the image is headed somewhere disclosure rules apply, keep the point below in mind: the file continues to identify itself as AI-generated through SynthID whatever the corner looks like now.
What clears a corner glyph from a generated image
The three we would use on a Gemini export, scored on the same 12 files as every tool on this site.
Swipe sideways for price, free export and platforms.
| # | Tool | Score | From | Free export | Best for | Runs on | Action |
|---|---|---|---|---|---|---|---|
| 1 |  Photoshop Generative FillPhotoshop | $9.99/mo | None | Photographers who cannot accept a soft patch | Win, Mac, Web, iOS | Visit Photoshop Generative Fill (opens in a new tab) | |
| 2 |  Dewatermark.aiDewatermark.ai | $6.99/mo | Three images a day, no account required | A free image fix in one minute | Web | Visit Dewatermark.ai (opens in a new tab) | |
| 3 |  Fotor Watermark RemoverFotor | $12.99/mo | Free downloads carry a Fotor watermark | Full-resolution stills on a small budget | Web, Win, Mac, iOS, Android | Visit Fotor Watermark Remover (opens in a new tab) |
One mark sells, the other accounts
The visible sparkle does marketing work. It labels free output as AI-made and as Gemini’s, and paid Google AI tiers drop it exactly the way Veo’s paid tiers drop their corner logo. One mark, one upsell, the standard arrangement across generators.
SynthID does the accountability work. Embedded across the pixels at generation, invisible, engineered to survive crops, filters and re-encoding, it lets platforms and detectors identify the image as AI-generated no matter what becomes of the sparkle. That layer is not a removal target for our tools or anyone else’s, and the AI watermark remover ranking covers what genuinely is.
Cosmetic cleanup, not anonymisation
On your own generations the visible mark is free-tier labelling applied to your own output. Removing it, or paying for the tier that skips it, is ordinary use of the product, and that is the scope of this guide.
It is not a route to presenting AI images as photography. SynthID still identifies the file, and disclosure rules in advertising, political content and news attach to the image rather than to the logo. Other people’s generations, and any attempt to launder them, sit outside what these tools are for.
Other generator marks worth understanding
Same tool, same method. Each guide opens with the official fix for that app.
Gemini watermark and SynthID questions
What actually removes the sparkle?
For images you generated: Dewatermark.ai detects the corner glyph and clears it free, twice a day in HD, while Photoshop’s Generative Fill handles marks that cross detail. Paid Google AI tiers generate without the visible mark in the first place.
Does subscribing stop it appearing?
Paid tiers generate images without the visible sparkle, which beats cleaning it off afterwards on every measure. The invisible SynthID layer remains on every plan, because it is provenance infrastructure rather than a free-tier restriction.
What is that icon in the corner?
Gemini’s visible AI mark, stamped into the corner of free-tier generations to label them as AI-made. It is an ordinary static corner mark, the easiest kind there is to remove, and the kind paid tiers simply never add.
Can the invisible layer be stripped?
No. SynthID is woven invisibly through the image at generation time and designed to survive editing, cropping, screenshots and re-encoding. No removal tool targets it, and since it is invisible it changes nothing about how your image looks or prints.
Will a platform still know the image is AI?
If it checks, yes. SynthID and provenance metadata continue to identify it, and services increasingly read both. Treat visible-mark removal as cosmetic tidying of your own work rather than as anonymisation, and follow disclosure rules wherever they apply to you.
Is any of this against the rules?
From your own generations, no. It is your output, and Google sells tiers that export without the mark. Removing marks from somebody else’s AI images, or stripping provenance to disguise AI content where disclosure is required, is the line this guide does not cross.
Does a screenshot avoid the whole problem?
It avoids nothing useful. A screenshot still contains the visible glyph unless you crop it out, and cropping costs you composition. SynthID is designed to survive screenshots as well, so the file remains identifiable. Screenshotting simply produces a smaller, lower-quality copy of the same situation.
Does the glyph appear on every image, or only some?
On free-tier generations it is consistent, though where it lands varies with aspect ratio and it occasionally falls somewhere it competes with your subject. Images arriving from the API rather than the consumer app follow different rules, so check a generation before assuming the whole batch is marked identically.
The cleanest Gemini image, fastest
Dewatermark.ai clears the Gemini glyph free, twice a day in HD. A paid Google AI tier is cleaner still, because the glyph is never applied.