What is the name of Google's text-to-image AI model?

Google's text-to-image AI model is Imagen.

Imagen was introduced by Google Brain in 2022 as a system that creates images from natural-language descriptions. It combines a language model's ability to interpret detailed prompts with diffusion-based image generation, allowing the system to turn phrases into pictures with specific subjects, styles, lighting, and compositions.

The original research emphasized the value of using a large language model trained on text to improve understanding of image prompts. Imagen's generation process uses cascaded stages: a base image is created at low resolution and then progressively enlarged and refined. Later releases expanded its capabilities, including improved detail, typography, editing, and image quality.

Imagen is a model family rather than one frozen product name; versions have included Imagen 2, Imagen 3, and Imagen 4. Stable Diffusion, Midjourney, and DALL-E are separate text-to-image systems from other organizations, so they are not alternative names for Google's model.