Interp3D: Correspondence-aware Interpolation for Generative Textured 3D Morphing

TL;DR

Interp3D tackles textured 3D morphing by integrating correspondences with a 3D diffusion prior in a training-free framework. It enforces a progressive alignment across semantic, structural, and texture dimensions, using semantic-aligned condition interpolation, SLAT-guided structure interpolation, and fine-grained texture fusion. A new dataset Interp3DData and comprehensive evaluations on fidelity, transition smoothness, and plausibility demonstrate superior performance over 3D baselines and 2D-to-3D adaptations. The method enables more faithful geometric evolution and consistent appearance across intermediate states, with practical implications for animation, editing, and digital content creation. The work provides a versatile design that can be integrated with multiple 3D backbones and offers a benchmark for textured 3D morphing research.

Abstract

Textured 3D morphing seeks to generate smooth and plausible transitions between two 3D assets, preserving both structural coherence and fine-grained appearance. This ability is crucial not only for advancing 3D generation research but also for practical applications in animation, editing, and digital content creation. Existing approaches either operate directly on geometry, limiting them to shape-only morphing while neglecting textures, or extend 2D interpolation strategies into 3D, which often causes semantic ambiguity, structural misalignment, and texture blurring. These challenges underscore the necessity to jointly preserve geometric consistency, texture alignment, and robustness throughout the transition process. To address this, we propose Interp3D, a novel training-free framework for textured 3D morphing. It harnesses generative priors and adopts a progressive alignment principle to ensure both geometric fidelity and texture coherence. Starting from semantically aligned interpolation in condition space, Interp3D enforces structural consistency via SLAT (Structured Latent)-guided structure interpolation, and finally transfers appearance details through fine-grained texture fusion. For comprehensive evaluations, we construct a dedicated dataset, Interp3DData, with graded difficulty levels and assess generation results from fidelity, transition smoothness, and plausibility. Both quantitative metrics and human studies demonstrate the significant advantages of our proposed approach over previous methods. Source code is available at https://github.com/xiaolul2/Interp3D.

Figures (12)

• Figure 1: We propose Interp3D, a training-free approach for continuous and plausible textured 3D morphing, consistently surpassing prior approaches with better structure fidelity, plausible appearance, and transition smoothness. Zoom in to check the details.
• Figure 2: Artifacts caused by missing correspondence alignments.
• Figure 3: Pipeline Overview. The left presents the overall framework. The right highlights component designs. Based on the 3D generation prior, the interpolation is progressively enhanced from three aspects: (a) Semantic-Aligned Condition Interpolation, (b) SLAT-Guided Structure Interpolation in structure generation, and (c) Fine-Grained Texture Fusion for appearance refinement.
• Figure 4: Qualitative Results. Our Interp3D achieves smooth and plausible 3D morphing.
• Figure 5: Visualized analysis for the effects of each component design.