Experience the cutting-edge image creation capabilities of Imagen 3, incorporating built-in protections and adhering closely to our core principles of product design. Imagen 3 consistently outperforms its counterparts in a variety of metrics, setting a new benchmark in image generation models. Just like Imagen 2, we integrate SynthID for watermarking AI-produced images.

Our design ethos is straightforward: You maintain full command of the creative journey from inception to completion. Should the initial image not quite meet your vision, communicate your desired adjustments to Gemini, and it will promptly furnish you with a new rendition.

In the upcoming days, we are introducing the capability to generate human images, initially accessible to our Gemini Advanced, Business, and Enterprise users, commencing in English. We have implemented technical enhancements to the product, refined evaluation sets, conducted red-team assessments, and established clear product guidelines.

Through Imagen 3, we have made remarkable strides in enhancing user satisfaction when crafting images of people. We refrain from supporting the generation of ultra-realistic recognizable individuals, portrayals of minors, or excessively graphic, violent, or explicit scenes. While not every image generated by Gemini may be flawless, we value and implement feedback from early users to drive continuous improvement. Our phased rollout will gradually expand accessibility to more users and languages.

Be among the initial cohort to enjoy these fresh attributes by exploring Gemini Advanced today or enrolling in Gemini for Workspace.