Google has announced that it will incorporate its advanced image generation technology, Imagen 3, into its Gemini AI model. This move also marks the resumption of Google’s ability to generate images of people, a feature it had paused earlier this year due to the creation of historically inaccurate depictions, such as Black founding fathers.
The introduction of Imagen 3 into Gemini promises more powerful image generation capabilities. Google is implementing new safety measures to address past issues. Google will not generate photorealistic images of people or create images involving children or identifiable individuals. Additionally, the company aims to prevent the depiction of excessively violent, gory, or sexual content.
The ability to create images that include people is limited to English prompts. It is only available to Gemini Advanced, Business, and Enterprise users. Furthermore, Google is expanding its "Gems" feature, which allows users to create domain-specific versions of Gemini. Initially previewed at Google I/O, the Gems feature is now accessible in 150 languages for the same group of paid users.
This update reflects Google’s cautious yet progressive approach to balancing innovation with ethical concerns, particularly in the realm of AI-generated content.