ControlNeXt: Powerful and Efficient Control for Image and Video Generation

TL;DR

ControlNeXt introduces a streamlined, resource-efficient approach for controllable image and video generation, significantly reducing training complexity and enabling seamless style modifications without extra training.

Contribution

It presents a simplified architecture with minimal additional costs, reduces learnable parameters by up to 90%, and introduces Cross Normalization for faster, stable training convergence.

Findings

Demonstrates robustness across various models and data types.

Achieves up to 90% reduction in learnable parameters.

Enables style alteration without additional training.

Abstract

Diffusion models have demonstrated remarkable and robust abilities in both image and video generation. To achieve greater control over generated results, researchers introduce additional architectures, such as ControlNet, Adapters and ReferenceNet, to integrate conditioning controls. However, current controllable generation methods often require substantial additional computational resources, especially for video generation, and face challenges in training or exhibit weak control. In this paper, we propose ControlNeXt: a powerful and efficient method for controllable image and video generation. We first design a more straightforward and efficient architecture, replacing heavy additional branches with minimal additional cost compared to the base model. Such a concise structure also allows our method to seamlessly integrate with other LoRA weights, enabling style alteration without the need for additional training. As for training, we reduce up to 90% of learnable parameters compared to the alternatives. Furthermore, we propose another method called Cross Normalization (CN) as a replacement for Zero-Convolution' to achieve fast and stable training convergence. We have conducted various experiments with different base models across images and videos, demonstrating the robustness of our method.