Beyond Temporal Smoothing: Spatial Energy Budgets Stabilize One-Step Diffusion Editing

2026-09-26

AI Summary

This study addresses the object distortion and background interference caused by spatial energy mismatch in single-step diffusion editing by proposing the BudEdit framework. This method introduces a pioneering spatial energy budget mechanism that converts the total energy of the residual field into an explicit budget, which is then precisely allocated to the editing regions via joint cross-attention. By controlling both the location and magnitude of energy injection, BudEdit fundamentally eliminates background drift, achieving stable training-free and inversion-free editing. Evaluated on the PIE-Bench benchmark, BudEdit comprehensively outperforms ChordEdit, improving background PSNR by 2.1 dB while reducing DINO and LPIPS scores by 31% and 36%, respectively, all with superior inference speed.

Abstract

One-step text-guided diffusion editing is efficient but prone to spatially misallocated updates that distort the edited object and alter the background. Existing methods often improve stability by averaging the editing field across timesteps. We instead identify spatial energy misallocation as a distinct and measurable failure mode: across two independent noise draws, the residual field is essentially unrepeatable, making the background field unreliable for direct transport, while its total energy still sets a usable magnitude for the draw at hand. BudEdit turns that magnitude into an explicit budget and reallocates it to edit-relevant regions selected jointly by residual energy and cross-attention, controlling where editing energy is spent rather than averaging over timesteps. The resulting training-free, inversion-free editor spends the budget on transport and reuses it to scale a correction in a lower-noise gated refinement. The budgeted injection field matches its prescribed budget exactly and vanishes on the identified background support, by construction. On PIE-Bench with SD-Turbo, BudEdit outperforms ChordEdit under each method's reported default settings on all 11 evaluated metrics, including a $2.1$\,dB gain in background PSNR, $31$\% lower DINO, and $36$\% lower LPIPS, while improving all five editing-quality metrics and reporting the lowest runtime in the comparison.

Problem

• one-step diffusion editing
• spatial energy misallocation
• background preservation
• text-guided editing

Innovation

• One-Step Diffusion Editing
• Spatial Energy Budget
• Training-Free Editing
• Cross-Attention Guidance
• Gated Refinement