Imagen
Google's text-to-image research model that generates photorealistic images from natural language descriptions.
About
Imagen is a text-to-image diffusion model developed by Google Research that converts written prompts into high-resolution photorealistic visuals. It pairs a large frozen language model for deep text understanding with a cascade of diffusion models that progressively upscale output from low to high resolution, reaching 1024 by 1024 pixels. The project introduced DrawBench, a benchmark used to compare text-to-image systems. Imagen is a research initiative and has not been released as a public product or API; the team cited concerns about bias and potential misuse as reasons for withholding a public demo. Access is limited to Google's internal research context.