FlashEdit: Decoupling Speed, Structure, and Semantics for Precise Image Editing

Text-guided image editing with diffusion models has achieved remarkable quality but often suffers from prohibitive latency. We introduce $\mathbf{\text{FlashEdit}}$, a real-time localized image editing framework for the standard inversion-based editing setting. Its efficiency and precision stem from three key innovations: (1) a $\mathbf{\text{Cycle-Consistent}}\mathbf{\text{One-Step}}\mathbf{\text{Inversion}}\mathbf{\text{(COSI)}}$ pipeline that encourages manifold-aligned one-step inversion through cycle consistency; (2) a $\mathbf{\text{Background}}\mathbf{\text{Shield}}\mathbf{\text{(BG-Shield)}}$ technique that improves preservation of non-edited regions via structural self-attention intervention; and (3) a $\mathbf{\text{Sparsified}}\mathbf{\text{Spatial}}\mathbf{\text{Cross-Attention}}\mathbf{\text{(SSCA)}}$ mechanism that promotes precise edits by suppressing semantic leakage. Experiments on PIE-Bench demonstrate a strong preservation-efficiency trade-off, with edits completed in under 0.2 seconds and an over 150$\times$ speedup over DDIM-based multi-step editing. Our code will be made publicly available at \url{https://github.com/JunyiWuCode/FlashEdit}.