Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

Imagen 2 - our most advanced text-to-image technology

Dec 15, 2023 - deepmind.google
Imagen 2 is Google's advanced text-to-image diffusion technology that can generate high-quality, photorealistic images in line with user prompts. It uses the natural distribution of its training data to create more lifelike images and is available for developers and Cloud customers via the Imagen API in Google Cloud Vertex AI. The technology is also being used in Google Arts and Culture's Cultural Icons experiment. Imagen 2 has improved image-caption understanding, more realistic image generation, fluid style conditioning, and advanced inpainting and outpainting capabilities.

The technology is designed to be responsible, with robust guardrails in place to mitigate potential risks. It is integrated with SynthID, a toolkit for watermarking and identifying AI-generated content. Safety measures have been put in place to limit problematic outputs and comprehensive safety filters are applied to avoid generating potentially problematic content. The technology is continuously evaluated for safety as its capabilities and launches expand.

Key takeaways:

  • Imagen 2 is an advanced text-to-image diffusion technology developed by Google, capable of generating high-quality, photorealistic images based on user prompts.
  • The technology has been improved to better understand the relationship between images and words, increasing its understanding of context and nuance. It has also been enhanced to generate more realistic images, particularly of hands and human faces.
  • Imagen 2 offers image editing capabilities like 'inpainting' and 'outpainting', allowing users to generate new content directly into the original image or extend the original image beyond its borders.
  • Google has implemented robust safety measures to mitigate potential risks and challenges of the technology, including watermarking and identifying AI-generated content, and applying safety checks to training data, input prompts, and system-generated outputs.
View Full Article

Comments (0)

Be the first to comment!