In this work, we propose imgs2imgs, a lightweight training-free pipeline for making visually consistent edits to multiview image sets. Our pipeline operates on an image-to-image basis, applying partial forward diffusion to the original input to preserve its core structure while facilitating localized edits during the guided denoising process. However, unlike standard image-to-image approaches that operate on a single view and often allow global drift from one view to the other, imgs2imgs performs localized edits across multiple photos of the same scene. Our method builds on depth-conditioned ControlNet diffusion models and introduces two key innovations. First, we incorporate a mask-guided editing mechanism. By inverting the segmentation mask fed to the DDIM sampler, we force the diffusion process to be confined only to the content within the object-of-interest’s boundaries, while leaving the background unchanged. Second, to maintain cross-view semantic alignment, we introduce a training-free feature sharing mechanism. During the first image passing, the reference, we create a catalog of indexes and values of the corresponding U-Net middle-block features. Then, during the target pass, we inject the feature for each target location into the U-Net’s middle block. Together, these techniques enable high-fidelity, targeted edits across multiview scenes. Achieving the structural preservation benefits of image-to-image partial diffusion, the conditioning guidance of ControlNet signals, and the cross-view alignment necessary for consistent multiview editing through feature sharing, all without requiring retraining or specialized inpainting models.