About RemoveWatermark.dev
A free, open-source, 100% client-side tool that removes the visible sparkle watermark from images generated by Google Gemini AI — instantly, privately, and without any quality loss.
When Google Gemini started adding a visible ✨ sparkle watermark to all its AI-generated images, it created friction for creators, students, researchers, and designers who needed clean images for presentations, mockups, and academic work.
Most "solutions" online were shady AI inpainting tools that guessed the pixels underneath — introducing artifacts and requiring image uploads to unknown servers. That didn't sit right.
Since Gemini uses a mathematically precise Alpha Compositing technique to embed the watermark, the original pixel values can be mathematically recovered with 100% accuracy — no guessing needed. This tool does exactly that, entirely inside your browser.
Google Gemini composites its watermark using standard alpha blending:
Since we know alpha and logo_value from reference images, we simply invert the equation:
This gives exact, lossless pixel restoration. No AI. No guessing. Pure math.
100% Private
Images never leave your browser. No server, no account, no sign-up.
Lossless Quality
Mathematical pixel restoration — zero compression, zero artifacts.
Instant & Free
No waiting, no payment. Works on desktop and mobile.
Open Source
Full source on GitHub. Inspect every line.
Auto-Detection
Detects 48 px or 96 px watermark size based on image resolution.
Bulk Processing
Upload multiple images and download them all as a ZIP.
Abhin Krishna
BTech Electronics & Biomedical Engineering (with Honours) · Minor in AI
Model Engineering College, Thrikkakara, Eranakulam, Kerala, India
Abhin is a second-year engineering student with a deep interest in AI, neurotechnology, Brain-Computer Interfaces, and open-source software. He builds tools that are private, fast, and free — because good software shouldn't cost or compromise your data.
RemoveWatermark.dev is one of several open-source projects he maintains alongside his academic research work in EEG, EMG, wearable tech, and machine learning.
Support the Project
This tool is free and will always stay free. If it saved you time, consider supporting on Patreon — it helps with hosting costs and keeps development going.