ObjectMorpher: 3D-Aware Image Editing via Deformable 3DGS Models

Yuhuan Xie; Aoxuan Pan; Yi-Hua Huang; Chirui Chang; Peng Dai; Xin Yu; Xiaojuan Qi

cs.CVMar 30, 2026

ObjectMorpher: 3D-Aware Image Editing via Deformable 3DGS Models

Authors: Yuhuan Xie, Aoxuan Pan, Yi-Hua Huang, Chirui Chang, Peng Dai, Xin Yu, Xiaojuan Qi

Organizations: The University of Hong Kong

Abstract

Achieving precise, object-level control in image editing remains challenging: 2D methods lack 3D awareness and often yield ambiguous or implausible results, while existing 3D-aware approaches rely on heavy optimization or incomplete monocular reconstructions. We present ObjectMorpher, a unified, interactive framework that converts ambiguous 2D edits into geometry-grounded operations. ObjectMorpher lifts target instances with an image-to-3D generator into editable 3D Gaussian Splatting (3DGS), enabling fast, identity-preserving manipulation. Users drag control points; a graph-based non-rigid deformation with as-rigid-as-possible (ARAP) constraints ensures physically sensible shape and pose changes. A composite diffusion module harmonizes lighting, color, and boundaries for seamless reintegration. Across diverse categories, ObjectMorpher delivers fine-grained, photorealistic edits with superior controllability and efficiency, outperforming 2D drag and 3D-aware baselines on KID, LPIPS, SIFID, and user preference.

Figures & tables

Figure 2: Overview of our image editing pipeline. The object is lifted from 2D pixels to high-fidelity 3DGS. Real-time editing with local rigidity is applied based on user input. The object is then repositioned, and a generative model refines the edits for harmonious results.

Figure 3: Comparison of graph connections based on Euclidean distance and geodesic distance.

Figure 4: Illustration of the training data preparation and the pipeline of our generative composition model.

Panel of Figure 5 — Figure 5: Qualitative comparisons. We show results on 8 subjects (rows) across 5 methods (columns). Our method (Ours) is the only one that faithfully follows the non-rigid user guidance while maintaining photorealism. 2D methods fail to perform the 3D-aware edit, producing results nearly identical to the origin.

Figure 6: Visual results of our user study. Our method is consistently preferred across all three metrics (Guidance Following, Style Consistency, and Identity Preservation) over all baselines.

Methods	LPIPS $\downarrow$	SIFID $\downarrow$	KID $\downarrow$	RI	MT $\downarrow$
DragGAN	0.550	16.091	-0.056	✓	$<$ 10s
ImageSculpting	0.178	14.372	-0.055	✓	$\sim$ 30m
DragDffusion ∗	0.117	6.573	-0.075	$\times$	$\sim$ 2m
DragAnything	0.655	22.476	-0.047	$\times$	$\sim$ 70s
Anydoor	0.173	9.512	-0.049	$\times$	$\sim$ 10s
InstantDrag	0.218	13.744	-0.042	$\times$	$\sim$ 10s

Table 1: Quantitative evaluation. RI represents real-time interaction and MT represents model inference time.( ∗ DragDiffusion often edits images with minimal modifications.)

Panel of Figure 7 — Figure 7: Ablation on 3D representations (3D GS vs. Mesh) and the use of Generative Composition (w/ GC vs. w/o GC).

Panel of Figure 8 — Figure 8: Ablation on the deformation constraints.

3D Shape Editing

3D shape editing changes the geometry, appearance, or parts of digital objects using spatial controls, prompts, images, or generative models.

See all 53 papers

3DMorph: Single-Image-Guided Local 3D Shape Editing and Morphing
Jun 5, 2026Tobias Preintner, Yunfei Deng, Phillip Müller +53D Shape Editing
Institute of Advanced Computer Science, Leiden University, Leiden, The Netherlands · BMW Group, Munich, Germany
PhyEdit: Towards Real-World Object Manipulation via Physically-Grounded Image Editing
Apr 8, 2026Ruihang Xu, Dewei Zhou, Xiaolong Shen +2
Zhejiang University, Hangzhou, China
EditVerse3D: High-Quality 3D Object Editing with Region-Aware Learning
Jul 8, 2026Youtan Yin, Yanning Zhou, Jiacheng Wei +63D Shape Editing 3D Scene Editing
College of Computing and Data Science, Nanyang Technological University · Tencent AIPD

Jun 5, 2026J/K move · Enter open · S save

3DMorph: Single-Image-Guided Local 3D Shape Editing and Morphing

Tobias Preintner, Yunfei Deng, Phillip Müller, Sebastian Illing+4

Institute of Advanced Computer Science, Leiden University, Leiden, The Netherlands · BMW Group, Munich, Germany

3D Shape Editing

Open paper AskPDF arXiv

Despite recent progress in 3D generation, intuitive editing of existing shapes remains limited. Unlike images, which benefit from well-established inpainting tools, general 3D objects such as meshes still lack simple and effective methods for local shape editing. Existing approaches are often global, domain-specific, require complex user interaction, or focus on appearance (color and texture) rather than geometry. We introduce 3DMorph, a training-free framework for single-image-guided local 3D shape editing and morphing. Given an edited image showing a desired shape modification, our method automatically localizes the relevant 3D region and transfers 2D modifications to 3D while preserving unmodified areas. 3DMorph also enables intermediate shape generation between the original and edited objects, facilitating design exploration. To benchmark editing quality, we introduce Delta3D, an image-guided local 3D editing benchmark with paired ground-truth edits. Experimental results show that 3DMorph translates intuitive 2D edits into 3D, outperforming state-of-the-art generative and editing methods.

Abstract

3D Shape Editing

Explore similar work

3DMorph: Single-Image-Guided Local 3D Shape Editing and Morphing

PhyEdit: Towards Real-World Object Manipulation via Physically-Grounded Image Editing

EditVerse3D: High-Quality 3D Object Editing with Region-Aware Learning