Frequently Asked Questions
Everything you need to know about removing Gemini watermarks from AI images.
About the Watermark
What is the Gemini watermark?
It's a sparkle (✦) logo automatically embedded in images generated by Google Gemini AI. It uses the Nano Banana and Nano Banana 2 (Imagen 3) image generation models. The logo appears semi-transparently in the bottom-right corner of generated images. Ultra subscribers and Google AI Studio users may be exempt from this watermark.
What's the difference between Nano Banana and Nano Banana 2?
Nano Banana is Google's first Gemini image generation model, while Nano Banana 2 is its successor built on Imagen 3 with improved image quality. Both use the same Gemini sparkle (✦) watermark, so this tool works with images from either model.
Where exactly is the watermark placed?
The sparkle logo is placed in the bottom-right corner. For images ≤ 1024px, it's 48×48px with a 32px margin. For images > 1024px, it's 96×96px with a 64px margin.
Using the Tool
How do I remove the watermark?
Visit the tool page, then drag and drop your image (or click to upload, or paste with Ctrl/Cmd+V). The watermark is automatically detected and removed in about 1 second. See the full step-by-step guide →
What image formats are supported?
PNG, JPG/JPEG, and WebP. The output is always a lossless PNG file. For best results, use PNG originals (JPEG compression can introduce minor artifacts).
Can I process multiple images at once?
Yes. You can upload or drag multiple files at once. All images are processed in parallel, and you can download them individually or all at once using "Save All."
Does it work on mobile?
Yes. The tool works in all modern mobile browsers including Chrome, Safari, Firefox, and Edge. No app installation needed.
What if the watermark isn't detected?
This can happen if the image has been heavily edited, cropped, or re-compressed before upload. For best results, use the original AI-generated image without modifications.
How It Works
How does the removal technology work?
The tool uses two techniques: NCC (Normalized Cross-Correlation) to precisely locate the watermark, and reverse alpha blending to mathematically recover the original pixels beneath it. This is not AI-based guessing — it's an exact mathematical reversal of the compositing process. Compare with other methods →
How accurate is the removal?
For PNG originals, the restoration is mathematically near-perfect. JPEG images may have minor artifacts due to compression, but results are still very clean. Always check the before/after slider to verify.
Why is reverse alpha blending better than inpainting?
Inpainting uses AI to guess what's beneath the watermark. Reverse alpha blending mathematically calculates the actual original pixels. The result is exact reconstruction rather than AI estimation. See a full comparison of methods →
Privacy & Legal
Are my images uploaded to any server?
No. All processing happens 100% in your browser using the Canvas API. Zero data is transmitted over the network. You can verify this by monitoring your browser's Network tab — no image-related requests are made.
Is it legal to remove AI watermarks?
Removing the visible watermark from images you generated yourself is generally not illegal. However, removing watermarks from others' copyrighted content may violate intellectual property laws. This tool is intended for use with your own AI-generated images only.
Does this remove SynthID?
No. SynthID is Google DeepMind's invisible watermark embedded at the pixel level during image generation. It's completely separate from the visible sparkle logo and is not affected by our tool. SynthID remains intact after processing.
Can the cleaned image still be detected as AI-generated?
Yes. While the visible sparkle logo is removed, the invisible SynthID watermark remains. Tools that detect SynthID will still identify the image as AI-generated.
Need More Detail?
If the short answers here are helpful, these pages expand on the exact workflow, the comparison of methods, and the difference between the visible logo and SynthID.