Gemini recently introduced new features, both in terms of customizing Gems and improving image generation using the Imagen 3 model. The new features, previewed at Google I/O, are now available to users on specific subscription plans, with the aim of improving the user experience and expanding the capabilities offered by the platform. One of the main new features is the ability to create custom Gems for Gemini Advanced, Business, and Enterprise users. Gems are configurable versions of Gemini, designed to act as virtual experts on specific topics. Users can customize their Gems by entering step-by-step instructions, giving them a name, and using them to get help with everything from programming to career planning. The feature saves time and streamlines complex processes, thanks to the ability for Gems to remember and apply a set of predefined instructions.
In parallel, the new image generation model, Imagen 3, has been introduced, which promises a significant improvement in the quality of the images produced. This update is available on Gemini, both in desktop and mobile versions, and is accessible in all supported languages. Imagen 3 allows you to generate high-quality images even from short text descriptions and allows the creation of images in a variety of styles, such as photorealistic landscapes or oil paintings. The model stands out for its advanced performance, complemented by protection and watermarking features through the use of SynthID.
The introduction of Imagen 3 brings with it further developments, such as early access to generating images of people for Gemini Advanced, Business, and Enterprise users. However, this feature is subject to strict ethical guidelines, which exclude the creation of photorealistic and identifiable images of individuals, as well as the depiction of minors or violent or inappropriate content. The goal is to ensure responsible and safe use of the technology, in line with Google’s design principles. New features will be released gradually, and future developments will depend on feedback from early access users.
