Unlike models that added image or audio understanding onto a text-only base, Gemini was trained jointly on text, images, audio, and video. It powers Google's consumer chat app, Search AI features, Android system integrations, and the Gemini API for developers.