The Google Gemini AI model is being used to remove watermarks from photos, which creates issues with digital rights and copyright. Social media users have found that Gemini may eliminate watermarks from stock photos, including those from Getty Images and other well-known media sources, despite Google’s efforts to enhance AI-generated content labelling.
Gemini 2.0 Flash can remove watermarks on photographs provided by Getty photographs and other popular stock media websites, as several users on Reddit and X have noted.
Google made its Gemini 2.0 Flash model’s image production feature more widely available last week, allowing the model to create and modify picture material natively. By all measures, it’s a potent ability. However, it also seems to lack many barriers. In addition to removing watermarks from pre-existing photos, Gemini 2.0 Flash will unreservedly produce images of celebrities and copyrighted figures.
In addition to eliminating watermarks from pre-existing images, the model’s text-to-image production function seems to have fewer restrictions and is said to be able to produce natively images of celebrities and other copyrighted content.
Gemini 2.0 Currently classified as “experimental” and “not for production use,” Flash’s image generating feature is limited to Google’s developer-facing products, such as AI Studio. Additionally, the model isn’t a flawless watermark remover. Certain semi-transparent watermarks and watermarks that cover a lot of a picture seem to be difficult for Gemini 2.0 Flash to handle.
In addition to removing watermarks, Gemini 2.0 Flash will try to fill in any gaps left by a watermark’s removal, as multiple Reddit and X users pointed out. While other AI-powered products also do this function, Gemini 2.0 Flash appears to be particularly adept at it and is available for free.
last week, the image generating capability of Gemini 2.0 Flash was made available for “experimental” and “not for production use” purposes. The internet giant’s platform for hosting AI developer tools, Google AI Studio, is currently the only way to access it.
Google DeepMind declared in 2024 that its watermarking tool for AI-generated text, SynthID Text, would be made publically accessible. It declared last month that it would start marking AI-manipulated photos with SynthID in its Photos app (Google Photos).
