Specialised tool · For Google Gemini images

Gemini Watermark Remover

Erase the Gemini sparkle mark from any AI-generated image in seconds. Works on the Gemini logo in screenshots too. Free, no signup, 5 removes a day.

Erase unwanted objects and clutter — before
Erase unwanted objects and clutter — after
ObjectPeopleWatermarkText

Before / After

Erase unwanted objects and clutter

What is the Gemini watermark?

Google Gemini attaches two layers of attribution to every image it generates. The first is a visible sparkle icon — a small, semi-transparent star usually tucked into the bottom-right corner. It lets anyone tell at a glance that the picture came from an AI model rather than a camera or artist. You will see it on outputs from the Gemini web app, the Gemini iOS and Android apps, and images rendered through the Gemini API when the default watermarking flag is on.

The second layer is SynthID, an invisible pixel-level signature spread across the whole image. Google designed SynthID to survive common edits — cropping, light colour grading, mild JPEG re-compression — so that downstream systems can still verify the image as AI-generated even after it has been shared, re-saved, and reposted. SynthID sits beneath the visible pixels and cannot be seen with the eye or removed with a brush.

MagicRemover targets the visible sparkle only. Brush over the star, tap Remove objects, and the AI inpainting model rebuilds the pixels where the mark was — sky, gradient, wall, skin, whatever the image calls for. The untouched pixels, including the SynthID signature living across them, are preserved byte-for-byte. If you need a clean-looking corner for personal use, this tool does that. If your goal is to defeat provenance detection, it will not — SynthID keeps working in every region you did not paint over.

How to remove a Gemini watermark in 3 steps

  1. 1

    Upload your Gemini-generated image

    Drag the image from your desktop into the upload card, or tap to browse. JPG and PNG both work, and anything wider than 1536 pixels is resized automatically — you do not need to downscale the file yourself. Images saved from the Gemini web app, the Gemini iOS app, and the Android Gemini app all carry the sparkle mark in the same corner, so any of those export paths work identically here.

  2. 2

    Brush over the sparkle mark or logo

    Pull the brush slider on the right panel to roughly the size of the mark — 25–45 pixels handles most corner sparkles; logos in screenshots usually need 60–100. Paint every pixel of the star or logo plus a small margin of 3–6 pixels around the edge so the AI has clean context to blend into. Undo and Clear are right there if you overshoot. The red overlay shows exactly which pixels will be rewritten.

  3. 3

    Download the clean image

    Tap Remove objects. In about 15 seconds the AI inpainting model analyses the surrounding pixels — sky, skin, gradient, texture — and rebuilds the masked region. Hit Download to save the result as PNG at the original resolution. Everything outside your brush stays byte-for-byte identical to the upload; only the sparkle area is rewritten.

Remove Gemini logos and SynthID marks

People often search for a Gemini logo remover and a Gemini watermark remover interchangeably, but the two elements are not the same. The sparkle watermark is the AI-origin indicator baked onto generated images. The Gemini logo, by contrast, is Google's brand mark — it appears in screenshots of the Gemini chat interface, in app-icon crops, and across product marketing. Both can be removed with the same brush-and-remove workflow, but the brush size tends to differ.

For the corner sparkle, a brush size of 25–45 pixels usually covers it cleanly. For a logo sitting in the middle of a screenshot, try 70–120 pixels. In both cases, paint a small margin around the shape — 3 to 6 extra pixels — so the inpainting model has fresh context to blend into. Zoom in if the mark is tiny.

A note on the phrase SynthID remover: no inpainting tool can reliably strip SynthID from the portions of the image you did not paint over. SynthID is an invisible watermark spread across every pixel, engineered to stay legible through everyday edits. MagicRemover rewrites only the pixels you brush — so the untouched 99% of the image still carries the SynthID signature, and Google's detector can still identify the photograph as AI-generated. Take that into account before any public posting where provenance matters.

The short answer is yes for personal use, with limits once the image leaves your hands. Google's terms for Gemini content give you reasonably broad rights over what you generate: you may use, modify, and share it. Nothing in the terms explicitly forbids editing the sparkle off an image you created yourself.

The caveats arrive when removal is part of misrepresenting origin. Stock photography marketplaces increasingly require explicit AI disclosure. News outlets, academic publications, and journalism contests have rules that either forbid AI-generated imagery or require labelling. The EU AI Act and similar regulations in the UK, US, and Asia are trending toward mandatory disclosure for synthetic media. Stripping the visible watermark does not exempt you from any of those requirements — the image is still AI-generated, and SynthID in the untouched regions still says so.

Common legitimate uses: personal wallpapers, keepsakes, internal mood boards, design mockups where the sparkle is visually distracting, portfolio pieces with a clear AI disclaimer elsewhere, or creative composites where you disclose the source. Uses to avoid: reselling a cleaned image as original photography or art, entering the image into a competition that bans AI content, uploading to a stock site without flagging AI origin, or using the cleaned image to deceive a viewer about who or what made it. Edit for your own workflow, disclose when the work leaves your hands, and MagicRemover is doing what it is meant for.

Gemini watermark vs other AI watermarks

Different generators take different approaches to attribution. Here is how Gemini sits alongside the rest:

Gemini (Google)

Visible: Sparkle icon, bottom-right corner

Invisible: SynthID across the whole image (robust)

Removable: Visible mark yes — invisible layer stays

DALL·E 3 (OpenAI / ChatGPT)

Visible: Rainbow strip bottom-right (API & ChatGPT Plus)

Invisible: C2PA metadata in file headers

Removable: Visible yes; metadata gone on re-save

Midjourney

Visible: Usually none on the rendered image

Invisible: C2PA metadata (rolling out)

Removable: No visible mark to remove

Stable Diffusion (open source)

Visible: None by default; depends on the UI

Invisible: Varies — most front-ends inject nothing

Removable: Usually already clean

Adobe Firefly

Visible: Watermark on free-tier exports only

Invisible: C2PA Content Credentials

Removable: Visible yes; credentials stripped on re-encode

MagicRemover can strip the visible portion for every provider in this table. The invisible layers behave as described — generally beyond the reach of brush-based inpainting. If transparency about AI origin matters in your use case, keep those layers alone and be upfront in your caption.

FAQ

Questions about removing the Gemini watermark

Related tools