Gemini's image generator might get rid of its biggest pain point

Watermarks on AI-generated images are something that frustrates users. Google seems to have taken note: a new item “Watermark settings” has appeared in its app’s code, thanks to which the visible Gemini watermark could finally be turned off.

What exactly was found in the app?

The discovery comes from an analysis of Google app version 17.43.11, in which editors of Android Authority stumbled upon a previously hidden “Watermark settings” menu. A string in the code suggests that the upcoming toggle will allow the logo to be removed from images.

Specifically, we are talking about the small Gemini icon in the bottom right corner of the image. When you create with ChatGPT, you won’t find any similar watermark on the resulting images, but unfortunately, you have to put up with it when using Google’s AI. Removing it then requires additional steps, typically cropping the image, or covering it in Photoshop if you cannot afford to cut off that part of the image.

When Google Vice President Josh Woodward, who is in charge of Gemini, publicly asked what bothered people about the assistant some time ago, removing the watermark was among the most frequent wishes. Woodward then promised that the team behind the AI assistant would consider the option.

You won’t completely hide the AI origin of the image anyway

Even if the visible logo disappears from the image, you still won’t completely hide the fact that it’s an artificial intelligence creation. The invisible SynthID watermark, which Google weaves directly into individual pixels using a complex mathematical structure, will take care of that – you won’t recognize it with the naked eye, and you won’t erase it by cropping or editing. Furthermore, the origin of the image is also monitored by the more common C2PA standard stored in metadata, but that is easily removed.

Removing the watermark, which users currently remove themselves using the aforementioned techniques, would therefore not significantly affect the risk of deepfakes or other misleading content.

Do watermarks on AI-generated images bother you?

Source: Android Authority