Hierarchical Text-Conditional Image Generation with CLIP Latents

TL;DR

This paper introduces a two-stage hierarchical model leveraging CLIP embeddings for text-conditional image generation, enhancing diversity, style preservation, and enabling zero-shot image manipulation.

Contribution

It proposes a novel two-stage approach combining a prior and decoder with CLIP embeddings, improving image diversity and enabling zero-shot language-guided edits.

Findings

Diffusion-based decoders outperform autoregressive ones in quality and efficiency.

Explicit image representation generation enhances diversity with minimal realism loss.

The model enables zero-shot, language-guided image manipulation.

Abstract

Contrastive models like CLIP have been shown to learn robust representations of images that capture both semantics and style. To leverage these representations for image generation, we propose a two-stage model: a prior that generates a CLIP image embedding given a text caption, and a decoder that generates an image conditioned on the image embedding. We show that explicitly generating image representations improves image diversity with minimal loss in photorealism and caption similarity. Our decoders conditioned on image representations can also produce variations of an image that preserve both its semantics and style, while varying the non-essential details absent from the image representation. Moreover, the joint embedding space of CLIP enables language-guided image manipulations in a zero-shot fashion. We use diffusion models for the decoder and experiment with both autoregressive and diffusion models for the prior, finding that the latter are computationally more efficient and produce higher-quality samples.