Imagen

lastmod 2022-08-17created 2022-05-252 edits

Google's text-to-image diffusion model

Features

• AI model that generates photorealistic images through a deep level of language understanding
• In human evaluations, generates images preferred over OpenAI's DALL-E 2
  • In terms of quality and image-text alignment
• Only the research paper and sample images are released
  • Decided not to release a publicly usable demo version due to ethical issues, etc.

See also

• DALL-E

Favorite site

• Imagen: Text-to-Image Diffusion Models