Saved in:
Bibliographic Details
Main Authors: Xie, Liangbin, Pakhomov, Daniil, Wang, Zhonghao, Wu, Zongze, Chen, Ziyan, Zhou, Yuqian, Zheng, Haitian, Zhang, Zhifei, Lin, Zhe, Zhou, Jiantao, Dong, Chao
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2504.00996
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • This paper introduces TurboFill, a fast image inpainting model that enhances a few-step text-to-image diffusion model with an inpainting adapter for high-quality and efficient inpainting. While standard diffusion models generate high-quality results, they incur high computational costs. We overcome this by training an inpainting adapter on a few-step distilled text-to-image model, DMD2, using a novel 3-step adversarial training scheme to ensure realistic, structurally consistent, and visually harmonious inpainted regions. To evaluate TurboFill, we propose two benchmarks: DilationBench, which tests performance across mask sizes, and HumanBench, based on human feedback for complex prompts. Experiments show that TurboFill outperforms both multi-step BrushNet and few-step inpainting methods, setting a new benchmark for high-performance inpainting tasks. Our project page: https://liangbinxie.github.io/projects/TurboFill/