VISIBLE
The sparkle logo
A small, semi-transparent four-point star in the bottom-right corner. It is the only mark the Media watermark setting controls and the only one remover tools target.
GEMINI AND NANO BANANA, OCTOBER 2026
Gemini watermark remover: turn off the sparkle, know what stays
Most people searching for a Gemini watermark remover want the little sparkle off a Nano Banana image. Since August 2026 Google ships its own off switch for that. What no setting or tool removes is SynthID, the invisible mark in the pixels and in Gemini's text.
Scan my text free Updated
gemini.google.com
Today
Write a caption for my lighthouse poster.
Message Gemini
MacBook Pro
9:41
ClaudeCleanLocal scanPASTED FROM GEMINI
SCANNING
…hidden characters
U+00A0no-break space1
U+200Bzero-width space1
U+202Fnarrow no-break space1
Hidden characters cleaned. SynthID text lives in word choice, and the image sparkle is a Gemini setting.
9:41
ClaudeCleanLocal scanPASTED FROM GEMINI
SCANNING
…hidden characters
U+00A0no-break space1
U+200Bzero-width space1
U+202Fnarrow no-break space1
Hidden characters cleaned. SynthID text lives in word choice, and the image sparkle is a Gemini setting.
Short answer
To remove the Gemini watermark from new images, open Settings on gemini.google.com, choose Media watermark and switch it off, then generate again. Remover tools only erase the visible sparkle from older images. Neither route removes SynthID or C2PA metadata, and Gemini text keeps its SynthID signal too.
THE OFFICIAL ROUTE
How to remove the Gemini watermark with Google's own setting
This is the clean way to lose the sparkle: no third-party tool, no blur in the corner, no upload. It takes about a minute.
- 01
Open Gemini on the web
Go to gemini.google.com and sign in with the account you make images with. Google's help center describes the web path below. - 02
Open Settings at the bottom left
Settings sits at the bottom of the left sidebar. Click it to open the settings list. - 03
Choose Media watermark
This is the setting Google added in August 2026. Its description says it only controls the visible watermark. - 04
Switch it off
Turn the setting off. Nothing changes for images you already made, and SynthID plus C2PA metadata keep being added. - 05
Generate the image again
Run the same prompt again. The new image comes back without the sparkle in the corner. Old downloads keep theirs, so regenerate anything you still need.
THE OFFICIAL ROUTE
Four clicks, then generate the image again
gemini.google.com
Make a poster of a lighthouse at sunset
ActivityPrivacyMedia watermarkHelp
MacBook Pro
When the setting will not show up
Google says that in India, South Korea and Vietnam only Google AI Ultra subscribers can change it; everyone else there gets the visible watermark automatically. Secondary reports from ExtremeTech and How2Shout add that work and school accounts do not get the toggle and that it starts switched on. TechCrunch reported a rollout over the coming days from August 14, so a missing switch can also mean it has not reached your account yet.
We're striking a balance here between creative control and safety
THREE MARKS, ONE IMAGE
Which watermarks are on a Gemini image?
The sparkle is only the part you can see. Google stacks two more signals underneath, and Gemini text has a fourth of its own.
IN THE PIXELS
SynthID
An invisible watermark woven into the image itself. Google says all media generated by its tools carries it, and its help center says it will usually survive re-scaling, re-coloring and compression.
IN THE FILE
C2PA metadata
Content Credentials attached to the file that record it was made with AI. Google's help center says Gemini Apps adds them to all AI media, and Google is open-sourcing Credentio, a C2PA library, per TechCrunch.
IN THE WORDS
SynthID in Gemini text
Text answers in the Gemini app and on the web carry a SynthID signal in their word choices, which DeepMind announced in 2024. Our SynthID guide explains how it works.
all media generated by Google's tools are embedded with our imperceptible SynthID digital watermark
X-RAY A GEMINI IMAGE
Three marks, and which ones each route leaves behind
9:41
Gemini
Make a poster of a lighthouse at sunset
Here is your lighthouse poster.
Ask Gemini
X-ray on- Visible sparkleBottom-right corner of the imagePresent
- SynthID watermarkSpread through the pixels, invisiblePresent
- Content Credentials (C2PA)Metadata inside the filePresent
As downloaded. Every image you save from the Gemini app starts with all three. Only one of them is visible.
Metadata is the fragile one. It sits in the file wrapper, so a screenshot or any tool that writes a brand new file can leave it behind, which is why the X-ray above marks it as often lost after a remover. SynthID lives in the pixels themselves and travels with them.
COPYING GEMINI TEXT?
Scan the Gemini answer before you paste it anywhere.
ClaudeClean works on text, not images. It finds zero-width spaces, no-break spaces and soft hyphens in Gemini answers and removes them in your browser. SynthID stays, and no detector result is promised.
- 300 rewrite credits a month
- 20,000 characters per request
- .txt, .docx and .pdf files
BY SURFACE
Which Gemini surfaces carry which mark?
Where you made the image decides whether there is a sparkle at all. SynthID is the one constant. Grey means Google has not said.
CHECK YOUR SURFACE
Which Gemini surfaces carry which mark
Visible sparkleSynthIDC2PA metadata
Gemini app, most countries
Gemini app in India, South Korea, Vietnam (no Ultra)
Google AI Ultra
Flow
Google AI Studio
Gemini API
Visible sparkleYour choice
Gemini app, most countries
Media watermark setting, live since mid-August 2026. Applies to new images only.
Cycling through every cell. Tap one to stop.
| Surface | Visible sparkle | SynthID | C2PA metadata |
|---|---|---|---|
| Gemini app, most countriesMost users | Your choice since Aug 2026 (Media watermark setting) | Yes, always | Yes, always |
| Gemini app in India, South Korea, Vietnam | Yes, unless you subscribe to Google AI Ultra | Yes, always | Yes, always |
| Google AI Ultra | No, removed Nov 2025 | Yes | Yes |
| Flow | Covered by the Aug 2026 toggle | Yes | Yes |
| Google AI Studio | No, removed Nov 2025 | Yes | Not stated by Google |
| Gemini API | Not stated in the docs | Yes, on all generated images | Not stated in the docs |
Swipe to compare →
Sources: Google help center (Manage watermark settings in Gemini Apps), Nano Banana Pro announcement, Gemini API image generation docs, TechCrunch on the August 2026 setting.
What developers report about the API
The Gemini API docs only say that all generated images include a SynthID watermark. Developer David Yates wrote in May 2026 that API images come back with no visible mark and that Flow was watermark-free by default at the time. Treat both as field reports, not Google policy.
HOW WE GOT HERE
When did Google start watermarking Gemini output?
Text came first, then the visible logo became optional tier by tier, and in October 2026 the detector went public.
May 14, 2024
SynthID reaches Gemini text
Google DeepMind announced it was expanding SynthID to watermark AI-generated text in the Gemini app and web experience, plus video from Veo.Oct 2024
SynthID Text goes open source
DeepMind released its text watermarking method for other developers, as MIT Technology Review reported. The detector key for Gemini itself stays with Google.May 20, 2025
SynthID Detector preview
Google opened a portal to journalists and researchers on a waitlist, covering image, audio, video and text. It said over 10 billion pieces of content had been watermarked by then.Nov 20, 2025
Nano Banana Pro, and the first sparkle-free tiers
Google kept the visible watermark for free and Google AI Pro users but removed it for Google AI Ultra subscribers and inside Google AI Studio.Aug 14, 2026
The Media watermark setting
Google let users switch off the visible watermark for Nano Banana, Omni and Lyria output in Gemini and Flow, with Search listed as coming soon, per TechCrunch. SynthID and C2PA were not affected.Oct 7, 2026
SynthID Detector opens to everyone
Anyone can now check images, video and audio at synthid.com, worldwide and in English, with OpenAI, NVIDIA and Kakao named as partners and Apple coming soon.
UNDER THE HOOD
How do Gemini watermark remover tools work?
The well-known open-source removers do not guess or inpaint. They undo the blend that put the logo there, which works because the logo is always the same.
HOW REMOVERS WORK
Reverse alpha blending, one pixel row at a time
Original pixels (what Gemini drew)
With the sparkle blended on top
After the remover solves for the original
Alpha per pixel (how strongly the logo covers it)
marked = a × logo + (1 − a) × original original = (marked − a × logo) ÷ (1 − a)
Clean restore. With the exact alpha map, the math gives back the original pixels, give or take rounding.
Gemini paints the white sparkle over the image with partial transparency. If you know the logo and the transparency (alpha) of each of its pixels, the blend formula can be solved backwards for the original. The GargantuaX README says the alpha map was captured by rendering the logo on a known solid background, and that the logo is 96 by 96 pixels with a 64-pixel margin on larger images, 48 by 48 with 32 pixels on smaller ones.
MIT LICENSE
GargantuaX/gemini-watermark-remover
About 5,600 stars and 870 forks when we checked in October 2026. It ships a web tool, a Chrome extension, a userscript and a command-line tool, processes images locally and states it does not remove invisible watermarks.
MIT LICENSE
hex-ci/gemini-watermark-remover-plus
A small fork that adds a Custom Area mode for marks in other spots. It runs with no backend, says your data stays in your browser and is meant for personal and educational use only.
WHERE IT RUNS
Local is the safe default
Both projects work on your own device. Plenty of hosted remover sites want an upload; read their privacy terms before you hand over a private image.
What works in your favor
- The setting leaves no artifacts. The image is generated without the logo, so there is nothing to subtract.
- The setting needs no extra tool. Nothing to install and nothing to upload.
- A remover handles old images. Useful for images made before the setting existed or on plans without it.
- A remover is exact in the ideal case. With the right alpha map and position, the original pixels come back.
What to watch out for
- The setting only covers new images. Old images keep their sparkle; you have to regenerate.
- The setting is not everywhere. India, South Korea and Vietnam need Google AI Ultra, and reports say work and school accounts are left out.
- A remover can leave a ghost. Yates reports misdetection can leave a negative mark, worst on white backgrounds.
- Neither touches SynthID. Google says so for the setting and the tools say so themselves; only the visible logo goes.
THE PART THAT STAYS
Does removing the sparkle remove SynthID?
No. Every route on this page leaves SynthID in place, and Google now lets anyone check for it.
AFTER THE REMOVER
The corner is clean. The pixels still answer.
- 1. Gemini image
- 2. Sparkle removed
- 3. Uploaded to a SynthID check
- 4. Result
Sparkle bottom-right
synthid.com
Check an image for SynthID
lighthouse-clean.png
Sparkle bottom-right
synthid.com
Check an image for SynthID
lighthouse-clean.png
A positive result in Gemini reads that all or part of the content was created or edited by Google's AI models, per the help center. The same page caps uploads at 100 MB, with video under 90 seconds and audio under an hour. The public site at synthid.com warns, as Search Engine Journal quoted, that it is not a general AI detector: it looks for Google's watermark and partners' marks only.
We do not cover stripping SynthID
This page explains what the marks are and what each route leaves behind. It does not give instructions for defeating SynthID. If the goal is to pass an AI image off as human-made, Google's policy and the EU AI Act both point the other way, as the section on rules below shows.
GEMINI TEXT
Gemini text watermark and hidden characters
Text has two separate problems. One is invisible characters you can see in a scan and delete. The other is SynthID, which no character tool touches.
GEMINI TEXT, PASTED
What a character scan finds, and what it can't
Workshop note
Workshop note
Our spring workshop runs on Saturday, April 18, from 10 to 2. Bring a laptop and one draft you want feedback on. Coffee is on us, and the room has plenty of outlets.
Looks normalMacBook Pro
U+00A0No-break spaceSpace
U+200BZero-width spaceDelete
U+202FNarrow no-break spaceSpace
U+00ADSoft hyphenDelete
SynthID text: unchanged. It lives in which words Gemini picked, so no character scan can see or remove it.
REMOVABLE
Invisible characters
No-break spaces, zero-width spaces, narrow no-break spaces and soft hyphens can ride along when you copy text out of a chat app, a browser or an editor. Google has not said Gemini adds any on purpose. They break word counts, search and line wrapping, and an invisible character detector shows them.
STAYS
SynthID text
DeepMind says it works best on longer responses, is less effective on factual prompts, and loses confidence when text is thoroughly rewritten or translated. Our Gemini AI detector guide covers who can check for it today.
This guide is about removal. If you want the mechanics of the watermark itself, read what SynthID is. If you want to know whether a piece of text came from Gemini, the Gemini AI detector guide is the better page. For the same cleanup on OpenAI output, see the ChatGPT watermark remover guide, and for the wider picture, AI text watermarking.
What the cleaner does with your Gemini text
Paste the answer into the ClaudeClean cleaner and the hidden-character scan runs in your browser. The optional rewrite sends text to OpenAI and is best effort. It does not detect or remove SynthID, and no detector result is promised.
SEE THE CLEANUP
Paste a Gemini answer, scan it, rewrite only if you choose
ClaudeClean treats any pasted text the same way, whether it came from Gemini, ChatGPT or Claude. It does not process images.
01
> source text
Paste or drop a file
Paste a Gemini answer or upload a .txt, .docx or .pdf. Up to 2,000 characters works without an account.
02
scan report
3 zero-width1 soft hyphencleanable signals2Scan locally
Zero-width characters, regular and narrow no-break spaces, soft hyphens and homoglyphs are flagged and removed in your browser. Your text is not stored.
03
before → after
clearusefulanswerfocusedhelpfulresponse
Copied ✓Optional rewrite
Light, Medium or Strong passes reword AI habits with a before and after diff. Rewriting is best effort and sends text to OpenAI.
PRICING
Clean the text free. Rewrite more with Pro.
The scan and hidden-character cleanup are free. Pro adds room for rewrites and files.
PRO
CLAUDECLEAN PRO
ClaudeClean Pro
$5/ month
Cancel anytime.
- 300 credits every month (≈300,000 characters of rewriting)
- 20,000 characters per request
- Clean .txt, .docx and .pdf files
- Light / medium / strong intensity
- Dashboard with history
Checking account…
Free: 10 credits a month with an account, or try 2,000 characters right here without one. Strong rewrites cost 2× credits.
THE RULES
Is it allowed to remove the Gemini watermark?
Removing the sparkle is now something Google offers itself. Where it gets risky is what you do with the image afterwards.
Google's Generative AI Prohibited Use Policy (last modified December 17, 2024) does not mention watermarks by name. It does prohibit circumventing abuse protections or safety filters, and it lists this under misleading content:
Misrepresenting the provenance of generated content by claiming it was created solely by a human, in order to deceive.
In the EU, Article 50 of the AI Act makes providers mark AI output in a machine-readable format so it is detectable as artificially generated. Separately, people who publish deepfakes must disclose that the content was artificially generated or manipulated, whatever happened to the watermark. Both remover projects carry their own disclaimers too.
Generally fine
- Using Google's setting. It is Google's own feature, built for creative control.
- Cleaning up your own design work. A poster, mockup or slide where the logo is in the way and nobody is misled.
- Stripping invisible characters from text. That is formatting cleanup, not provenance.
- Disclosing AI use anyway. A caption or credit line keeps everything above board.
Where it goes wrong
- Passing an image off as a real photo. Claiming something is human-made to deceive is exactly what the policy bans.
- Publishing deepfakes in the EU without a label. Article 50 requires the disclosure regardless of watermarks.
- Breaking school or client rules. A missing sparkle does not change what you have to declare.
- Uploading private images to unknown sites. Hosted removers may keep what you send them.
Not legal advice
Rules differ by country, platform and contract. If an image is going into advertising, news or anything with legal weight, check the terms that apply to you.
PICK YOUR ROUTE
Which Gemini watermark remover route fits your case?
Short answers for the situations people search for most.
I want the sparkle off the images I make from now on
Switch Media watermark off in Gemini's settings and regenerate. It is free, official and leaves no artifacts.
I have older images that still show the sparkle
Regenerating with the setting off gives the cleanest result. If you cannot regenerate, a local reverse-alpha tool can erase the logo, with a small risk of a ghost outline on light backgrounds.
I am in India, South Korea or Vietnam
Google limits the setting there to Google AI Ultra subscribers, and Ultra images skip the visible mark since November 2025. On other plans the sparkle stays on by design.
I build with the Gemini API or AI Studio
Google removed the visible mark in AI Studio in November 2025. The API docs only promise SynthID on every image. Either way, plan for SynthID being there.
I need the image to look human-made
No route on this page gets you there. SynthID stays in the pixels, the public detector can read it, and claiming human authorship to deceive breaks Google's policy.
My problem is Gemini text, not an image
Clean the invisible characters with the ClaudeClean cleaner and rewrite in your own words if that is the goal. The AI text cleaner guide covers the details, and if you need to show your work is yours, see how to prove you didn't use AI.
SOURCES
Where these facts come from
Google's own pages first, then the reporting and the open-source projects. Checked October 10, 2026.
Google Help Center
Manage watermark settings in Gemini Apps
Media watermark path, India, South Korea and Vietnam limit, and confirmation that SynthID and C2PA stay.
View sourceGoogle Help Center
Verify Google AI-generated images and videos with SynthID
File limits, about 10 image checks per 24 hours, robustness to re-scaling and compression, result wording.
View sourceTechCrunch
Google will now allow users to remove visible watermark from its AI generations
Aug 14, 2026 report: Nano Banana, Omni and Lyria in Gemini and Flow, rollout timing, Woodward quote, Credentio.
View sourceNano Banana Pro announcement
Nov 20, 2025: SynthID on all media, visible mark kept for free and Pro, removed for Ultra and AI Studio.
View sourceSynthID Detector opens to the public
Oct 7, 2026: synthid.com for images, video and audio, partners, 180 billion items and 1 million daily checks.
View sourceGoogle DeepMind
Watermarking AI-generated text and video with SynthID
May 14, 2024: SynthID in Gemini app text, where it works best and what weakens it.
View sourceGoogle AI for Developers
Gemini API image generation docs
States that all generated images include a SynthID watermark.
View sourceGitHub
GargantuaX/gemini-watermark-remover
Reverse alpha blending, logo sizes and margins, local processing, no effect on invisible watermarks.
View sourceGitHub
hex-ci/gemini-watermark-remover-plus
Fork with Custom Area mode, browser-only processing, personal and educational use notice.
View sourceDavid Yates
Gemini watermark removal
May 2026 field notes on API and Flow output and on ghost marks after reverse alpha blending.
View sourceGenerative AI Prohibited Use Policy
Provenance misrepresentation and safety-filter circumvention clauses.
View sourceEU AI Act
Article 50: Transparency obligations
Machine-readable marking for providers and deepfake disclosure for deployers.
View sourceFAQ
Gemini watermark remover FAQ.
Use Google's own setting. On gemini.google.com, open Settings at the bottom left, choose Media watermark and switch it off, then generate the image again. It only affects new images and only the visible sparkle. In India, South Korea and Vietnam it is limited to Google AI Ultra subscribers.
No. Google says the Media watermark setting does not change SynthID or C2PA metadata, and the open-source remover tools state they do not touch invisible watermarks. A SynthID check can still report the image as made with Google AI.
Images from Google AI Ultra subscribers and from the Google AI Studio developer tool, per Google's November 2025 Nano Banana Pro post, plus any image made after you switch Media watermark off. Every one of them still carries SynthID.
The popular open-source ones use reverse alpha blending. They know the exact logo and how transparent each of its pixels is, so they solve the blend formula for the original pixels. It runs in your browser and can leave a faint ghost if the alpha map or position is off.
Google's Generative AI Prohibited Use Policy does not mention watermarks by name, and Google now offers an off switch for the visible one. The policy does ban misrepresenting provenance by claiming content was made solely by a human in order to deceive.
Yes. Google DeepMind said in May 2024 that SynthID was expanding to text in the Gemini app and web experience. The signal sits in word choice, so there is no character to delete. Thorough rewriting or translation reduces it, by Google's own account.
Not from images. ClaudeClean works on text: it strips invisible characters such as zero-width and no-break spaces from pasted Gemini answers in your browser, with an optional rewrite. It does not remove or detect SynthID and promises no detector result.
Yes. Google opened its SynthID Detector to the public on October 7, 2026 at synthid.com, and the Gemini app can check uploaded images and videos. Google's help center allows about 10 image checks per rolling 24 hours in the app.