CTRLorALTer: Conditional LoRAdapter for Efficient 0-Shot Control & Altering of T2I Models

TL;DR

This paper introduces LoRAdapter, a novel, efficient method for zero-shot control of text-to-image diffusion models, enabling detailed style and structure conditioning during image generation.

Contribution

LoRAdapter unifies style and structure conditioning with a new conditional LoRA block, improving zero-shot control in text-to-image models.

Findings

Outperforms recent state-of-the-art methods

Enables fine-grained control during generation

Architecture-agnostic and efficient approach

Abstract

Text-to-image generative models have become a prominent and powerful tool that excels at generating high-resolution realistic images. However, guiding the generative process of these models to consider detailed forms of conditioning reflecting style and/or structure information remains an open problem. In this paper, we present LoRAdapter, an approach that unifies both style and structure conditioning under the same formulation using a novel conditional LoRA block that enables zero-shot control. LoRAdapter is an efficient, powerful, and architecture-agnostic approach to condition text-to-image diffusion models, which enables fine-grained control conditioning during generation and outperforms recent state-of-the-art approaches.