Google has begun rolling out its latest “experimental” features of the Gemini 2.0 Flash AI model to developers globally, raising concerns over its capabilities. The AI model can not only generate images from text prompts but also conversationally edit existing ones.
One notable ability is its precision watermark removal, which can clean up even complex Getty Images stamps and SynthID marks without altering the image. However, this feature may be open to abuse as it also allows for the addition of recognizable images of real people, such as Elon Musk, in photos.
The Gemini 2.0 Flash AI model is available through Google’s AI Studio for developers, but its lack of guardrails raises questions about potential misuse. The company has not yet responded to queries regarding protections against watermark removal or other features like image manipulation.
While some tools, like Watermark Remover.io, already offer similar functionality, Gemini 2.0 Flash appears to have a competitive edge in this area. However, the availability of these features to everyone remains uncertain until Google provides further clarification on its usage guidelines and safeguards.
Source: https://www.theverge.com/news/631203/google-gemini-flash-2-native-image-generation-watermark-removal