Ready for another bleak application of generative AI?

I’m unsure if I am, but here we are with Gemini’s ‘Nano Banana’. This name is a playful twist on the major update to Gemini’s AI image generation features, referred to as Gemini 2.5 Flash.

Previously, Google introduced native image editing capabilities to the Gemini app, and this update enhances that model, allowing for visually consistent variations of characters or images with perhaps disturbing accuracy. While it’s advised not to upload your likeness into any generative AI model, this update shows how easy it is to envision yourself as different personas or characters.

Google describes how the Gemini app enables users to combine photos—whether to place themselves alongside pets or experiment with backgrounds—all while maintaining their original features. Impressively, this update has achieved a top position on the LMArena’s image editing model leaderboard.

In addition to assembling photos and creating outfit changes for subjects, the Gemini app introduces multi-turn editing capabilities. Users can capture an empty room and ask the app to fill it in with details over a few prompts without losing coherence in the visuals. This incremental editing ability clearly separates it from its competitors.

Despite its smart functionality, the app requires refinement, as it has struggled with basic tasks like cropping images accurately to specific aspect ratios. For example, it stated: “not capable of performing precise edits like cropping an image to a specific aspect ratio.”

While the Gemini app can help create amusing selfies or assist in interior decorating, it also raises ethical concerns by making it possible to alter images of deceased public figures, which can lead to outright misrepresentation.

Each image generated through the Gemini app includes a watermark indicating its AI origin, although this is subtle enough to be overlooked or removed by dishonest individuals using other editing tools. Moreover, Google mentions that all images have an ‘invisible SynthID digital watermark’ to help identify AI-generated content, although access to this is currently limited to select partners and early testers.

Ultimately, the ‘Nano Banana’ update and the rapid advancement of generative AI technology provoke serious apprehension about privacy and personal identity exposure in the digital age.