Gemini can now create personalized AI images by digging around in Google Photos

Perfectly imperfect

Google notes that the new feature is still evolving, so it might not always choose the right images. If that happens, you may want to check the sources list to see what went wrong. It will list the images referenced in the prompt, and you can also ask Gemini in a follow-up prompt about the images it chose. Manually selecting photos with the plus button in Gemini can help address these shortcomings.

While Nano Banana 2 can now peruse your Google Photos library when generating an image, Google stresses that it’s not retaining that data for training. The distinction between using your data in a prompt and training AI with it can be confusing, but Google says it doesn’t use any images from your library in training. However, it does use the inputs (what you type) and outputs (what the model does as a result) to improve AI products. That can still include personal data about you, but it’s not the same as pouring all your photos into…