Hierarchical Text-Conditional Image Generation with CLIP Latents

This paper introduces a hierarchical approach to text-conditional image generation. It leverages the CLIP (Contrastive Language-Image Pre-training) model to encode text descriptions into latent space and then uses a diffusion model to generate images based on these CLIP latents. This allows for generating high-quality images that align well with the given text prompts, by breaking down the image generation process into multiple stages.