SwiftPie: Lightning-fast Subject-driven Image Personalization via One step Diffusion

2026-05-02

AI Summary

Existing image personalization methods rely on time-consuming fine-tuning or multi-step denoising, hindering real-time interaction. This work proposes SwiftPie, the first high-fidelity, subject-driven personalized diffusion model capable of one-step generation. Its core innovations include a dual-branch identity injection mechanism and a mask-guided rescaling strategy, which efficiently integrate identity features with text prompts within a single denoising step. Experimental results demonstrate that SwiftPie achieves identity fidelity and prompt alignment comparable to multi-step approaches while significantly accelerating generation speed, thereby enabling high-quality, real-time personalized image synthesis.

Abstract

Diffusion models have achieved remarkable success in high-quality image synthesis, sparking interest in image-guided generation tasks such as subject-driven image personalization. Despite their impressive personalization results, existing methods typically rely on computationally intensive fine-tuning, iterative optimization, or multi-step denoising processes, which significantly hinder their deployment and interactive capability in real-time applications. In this work, we present SwiftPie, the first one-step diffusion image personalization tool that enables lightning-fast generation of personalized images. SwiftPie introduces a novel dual-branch identity injection mechanism that effectively integrates subject identity into a one-step diffusion model. In addition, we incorporate a mask-guided rescaling strategy to further enhance subject contextualization within a single diffusion step. Extensive experiments demonstrate that SwiftPie not only delivers superior image personalization speed but also achieves comparable performance with multi-step approaches in both identity fidelity and prompt alignment. This work opens new opportunities for real-time, high-quality personalized image generation, paving the way for interactive visual synthesis.

Problem

• subject-driven image personalization
• diffusion models
• real-time generation
• one-step diffusion
• image synthesis

Innovation

• one-step diffusion
• subject-driven personalization
• dual-branch identity injection
• mask-guided rescaling
• real-time image generation