Overview
Imagen 3 was Google DeepMind’s text-to-image model for generating realistic images from prompts. Google documentation described text-only input, multiple output images, aspect-ratio controls, safety settings, and SynthID watermarking, with access through the Gemini API and Vertex AI. Imagen 3 was a proprietary hosted model and had no official source repository for the product itself. Google’s current Gemini API documentation states that Imagen 3 has been shut down and directs developers toward newer Gemini image models, so this directory entry is marked dead rather than active.