Google's Imagen 3 was initially available through the company's AI Test Kitchen in the US, but starting today, all Gemini users can access the text-to-image model. Google says the new AI image generator model can offer more photorealism and produce fewer artifacts than its predecessor.