Google upgrades Gemini with Imagen 3 AI model for image generation; see details

Google has announced a major update to its AI model Gemini, incorporating its latest image generation model, Imagen 3, to power the visual capabilities of the Gemini chatbot. This upgrade, revealed on Wednesday, aims to boost Gemini’s image-generation features, making the advanced tools available to all users, including those on the free tier. Developers using the Gemini API will also be able to access Imagen 3 for building applications and experiences with enhanced visual capabilities.

Imagen 3, Google’s latest AI image generator, offers superior photorealism, improved adherence to prompts, and fewer unintended elements in generated images. In a post on X (formerly known as Twitter), Google confirmed that all users of the Gemini app can now create images with this advanced model, and early tests have shown that it outperforms comparable models like Meta AI. In a side-by-side comparison, Gemini and Meta AI were tasked with creating an image of a golden retriever on a train. While both missed some elements, Gemini’s Imagen 3 rendered more details accurately, at a higher resolution of 2048 x 2048, compared to Meta’s 1280 x 1280.

Imagen 3 supports a variety of artistic styles, from photorealism to textured oil paintings and even claymation. Additionally, users can specify a camera style for the image, such as a Nikon DSLR, a GoPro, or a wide-angle lens. This flexibility allows for creative freedom across different visual aesthetics, catering to both casual users and professional needs.

Security is a significant focus with this update. To combat the misuse of AI-generated content, Google has integrated SynthID technology, which embeds an invisible watermark directly into the pixels of each image. This AI label cannot be removed or cropped out, even in screenshots, and is designed to prevent the spread of deepfakes and other unauthorised manipulations.

The Imagen 3 model is now available within the Gemini app and API, making it easier than ever for developers and users alike to explore and leverage Google’s latest advances in AI image generation.