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:this https URL
View on arXiv@article{xie2025_2504.00996,
title={ TurboFill: Adapting Few-step Text-to-image Model for Fast Image Inpainting },
author={ Liangbin Xie and Daniil Pakhomov and Zhonghao Wang and Zongze Wu and Ziyan Chen and Yuqian Zhou and Haitian Zheng and Zhifei Zhang and Zhe Lin and Jiantao Zhou and Chao Dong },
journal={arXiv preprint arXiv:2504.00996},
year={ 2025 }
}