Continuous Control of Editing Models via Adaptive-Origin Guidance

TL;DR

This paper introduces Adaptive-Origin Guidance (AdaOr), a novel method that enables smooth, continuous control over text-guided diffusion-based image and video editing by interpolating guidance origins, improving edit consistency and flexibility.

Contribution

AdaOr adjusts guidance origins with identity-conditioned predictions, allowing fine-grained, continuous control in diffusion-based editing without additional training or datasets.

Findings

Provides smoother, more consistent editing control

Enables fine-grained adjustment of edit strength

Works effectively on both image and video editing tasks

Abstract

Diffusion-based editing models have emerged as a powerful tool for semantic image and video manipulation. However, existing models lack a mechanism for smoothly controlling the intensity of text-guided edits. In standard text-conditioned generation, Classifier-Free Guidance (CFG) impacts prompt adherence, suggesting it as a potential control for edit intensity in editing models. However, we show that scaling CFG in these models does not produce a smooth transition between the input and the edited result. We attribute this behavior to the unconditional prediction, which serves as the guidance origin and dominates the generation at low guidance scales, while representing an arbitrary manipulation of the input content. To enable continuous control, we introduce Adaptive-Origin Guidance (AdaOr), a method that adjusts this standard guidance origin with an identity-conditioned adaptive origin, using an identity instruction corresponding to the identity manipulation. By interpolating this identity prediction with the standard unconditional prediction according to the edit strength, we ensure a continuous transition from the input to the edited result. We evaluate our method on image and video editing tasks, demonstrating that it provides smoother and more consistent control compared to current slider-based editing approaches. Our method incorporates an identity instruction into the standard training framework, enabling fine-grained control at inference time without per-edit procedure or reliance on specialized datasets.