Conditional Text-to-Image Generation with Reference Guidance

TL;DR

This paper introduces reference-guided conditioning for text-to-image diffusion models, improving the rendering of specific subjects like text and extending capabilities to multilingual and logo generation.

Contribution

It proposes expert plugins that incorporate visual reference conditions into diffusion models, enhancing accuracy and generalization for specialized text and image synthesis tasks.

Findings

Superior results on all tasks compared to existing methods

Efficient plugins with only 28.55M parameters

Extended capabilities to multilingual and non-English text generation

Abstract

Text-to-image diffusion models have demonstrated tremendous success in synthesizing visually stunning images given textual instructions. Despite remarkable progress in creating high-fidelity visuals, text-to-image models can still struggle with precisely rendering subjects, such as text spelling. To address this challenge, this paper explores using additional conditions of an image that provides visual guidance of the particular subjects for diffusion models to generate. In addition, this reference condition empowers the model to be conditioned in ways that the vocabularies of the text tokenizer cannot adequately represent, and further extends the model's generalization to novel capabilities such as generating non-English text spellings. We develop several small-scale expert plugins that efficiently endow a Stable Diffusion model with the capability to take different references. Each plugin is trained with auxiliary networks and loss functions customized for applications such as English scene-text generation, multi-lingual scene-text generation, and logo-image generation. Our expert plugins demonstrate superior results than the existing methods on all tasks, each containing only 28.55M trainable parameters.