Gemini can now create AI images using your photo library
What's the story
Google has integrated its image-generation model with Google Photos, as part of the "personal intelligence" feature in Gemini. The update allows users to generate personalized AI images by using their own photos and associated labels. The move is aimed at making the process of creating unique content more efficient and accurate for users.
Enhanced functionality
How it works
The integration of personal intelligence into Google's Nano Banana 2, one of the top AI image generators, simplifies the process of creating new AI content. Users can now let the image bot access their photo content for more personalized results. This is especially useful as adding more personal data to an AI prompt usually leads to better outputs.
Practical examples
Linking photos for better results
Google has shared a few examples showing how linking Nano Banana to Photos can improve results. For instance, if you want a claymation image of your family enjoying their favorite activity, Gemini will use the labels you've added in Google Photos to understand "family." The content of the images can also help it figure out what a "favorite activity" is. This way, users don't have to provide as much context in their prompts.
Room for improvement
Feature is still a work in progress: Google
Google has acknowledged that the new feature is still a work in progress and may not always select the right images. In such cases, users can check the sources list to understand what went wrong. The list will show the images referenced in the prompt and users can ask Gemini about its choices using follow-up prompts. Manually picking photos with a plus button in Gemini can also help mitigate these issues.
Privacy assurance
Your data isn't being used to train Gemini
Despite the new capability, Google has assured users that their data isn't being used for training purposes. The tech giant makes a clear distinction between using your data in a prompt and training AI with it. While inputs and outputs are used to improve AI products, no images from your library are included in the training process.