Architecture
Zero-Shot Text-to-Image Generation
OpenAI·February 24, 2021
Aditya Ramesh, Mikhail Pavlov, Gabriel Goh
View on arXivTL;DR
The original DALL·E generates images from text captions with a Transformer over image tokens.
Why it matters
Made text-to-image generation a mainstream idea and previewed the generative-image boom.