Google’s Imagen 2 is an advanced text-to-image technology that generates high-quality and photorealistic images based on user prompts. It utilizes the natural distribution of training data instead of pre-programmed styles, resulting in more lifelike images. The technology has been integrated into Google Cloud Vertex AI and is also being used by the Google Arts and Culture team. Imagen 2 has improved understanding of context and nuance by enhancing image-caption pairings in its training dataset. It also generates more realistic images, addresses challenges with rendering human faces and hands, and allows for style conditioning and image editing capabilities. Safety measures, such as robust guardrails and safety filters, have been implemented to mitigate potential risks.
https://deepmind.google/technologies/imagen-2/