Title:Exploring Optimal Latent Trajetory for Zero-shot Image Editing

Abstract:Editability and fidelity are two essential demands for text-driven image editing, which expects that the editing area should align with the target prompt and the rest should remain unchanged separately. The current cutting-edge editing methods usually obey an "inversion-then-editing" pipeline, where the source image is first inverted to an approximate Gaussian noise ${z}_T$, based on which a sampling process is conducted using the target prompt. Nevertheless, we argue that it is not a good choice to use a near-Gaussian noise as a pivot for further editing since it almost lost all structure fidelity. We verify this by a pilot experiment, discovering that some intermediate-inverted latents can achieve a better trade-off between editability and fidelity than the fully-inverted ${z}_T$. Based on this, we propose a novel editing paradigm dubbed ZZEdit, which gentlely strengthens the target guidance on a sufficient-for-editing while structure-preserving latent. Specifically, we locate such an editing pivot by searching the first point on the inversion trajectory which has larger response levels toward the target prompt than the source one. Then, we propose a ZigZag process to perform mild target guiding on this pivot, which fulfills denoising and inversion iteratively, approaching the target while still holding fidelity. Afterwards, to achieve the same number of inversion and denoising steps, we perform a pure sampling process under the target prompt. Extensive experiments highlight the effectiveness of our ZZEdit in diverse image editing scenarios compared with the "inversion-then-editing" pipeline.