Google’s Gemini AI is really good at watermark removal
Mar 17, 2025 - theverge.com
Google's Gemini 2.0 Flash AI model is being rolled out to more developers globally, featuring advanced image generation and editing capabilities. Users have discovered that it can remove watermarks from photos with precision, raising concerns about copyright protection. While tools like Watermark Remover.io and Google's own 2017 algorithm have addressed similar issues, Gemini 2.0 Flash appears more adept at handling complex watermarks, such as those from Getty Images. After removing a watermark, the AI adds a SynthID mark, indicating the image was edited with AI, although this too can be removed using AI tools.
Additionally, Gemini 2.0 Flash can insert recognizable images of real people, like Elon Musk, into photos, a feature not available in the full Gemini model. Currently, these image editing features are accessible only to developers via AI Studio, limiting public access and potential misuse. Google has yet to respond to inquiries about protective measures against misuse, such as watermark removal.
Key takeaways:
Google's Gemini 2.0 Flash AI model can remove watermarks from images with precision, raising concerns about image protection.
The AI model can generate images from text prompts and conversationally edit images, including adding recognizable images of real people.
After removing a watermark, the AI adds a SynthID mark, indicating the image was edited with AI, but this mark can also be removed using AI tools.
Currently, the advanced image editing features of Gemini 2.0 Flash are available only to developers through AI Studio, limiting public access and potential misuse.
