Hierarchical text-conditional image generation with CLIP latents

Read on openai.com