3D editing refers to the ability to apply local or global modifications to 3D assets. Effective 3D editing requires maintaining semantic consistency by performing localized changes according to prompts, while also preserving local invariance so that unchanged regions remain consistent with the original. However, existing approaches have significant limitations: multi-view editing methods incur losses when projecting back to 3D, while voxel-based editing is constrained in both the regions that can be modified and the scale of modifications. Moreover, the lack of sufficiently large editing datasets for training and evaluation remains a challenge. To address these challenges, we propose a Beyond Voxel 3D Editing (BVE) framework with a self-constructed large-scale dataset specifically tailored for 3D editing. Building upon this dataset, our model enhances a foundational image-to-3D generative architecture with lightweight, trainable modules, enabling efficient injection of textual semantics without the need for expensive full-model retraining. Furthermore, we introduce an annotation-free 3D masking strategy to preserve local invariance, maintaining the integrity of unchanged regions during editing. Extensive experiments demonstrate that BVE achieves superior performance in generating high-quality, text-aligned 3D assets, while faithfully retaining the visual characteristics of the original input.
Figures & tables
Figure 2: Dataset Construction pipeline and Sub-Task Distribution
Figure 3: Data Construction Pipeline
Figure 4: Overview of our method. Structure Editing: The Flow Edit Transformer modifies the input 3D asset’s sparse structure based on a text prompt and a render image from original 3D asset. Structured Latent Editing: The Sparse Flow Edit Transformer enables fine-grained material and texture modifications.
Figure 5: The network structures for generation editing.
Figure 6: Editing Preservation Mask
Method
CD ↓
SSIM ↑
LPIPS ↓
FID ↓
FVD ↓
DINO-I ↑
CLIP-T ↑
Vox-E [ 61 ]
/
0.539
0.346
217.9
8590
0.371
0.051
Tailor3D [ 55 ]
0.067
0.751
0.198
146.3
4743
0.633
0.21
TRELLIS [ 81 ]
0.063
0.865
0.225
140.7
3986
0.910
0.25
Hunyuan [ 68 ]
0.021
0.853
0.087
119.83
1520
0.850
0.269
Ours (full)
0.013
0.960
0.039
28.9
1139
0.960
0.287
w/o MASK
0.019
0.824
0.048
49.2
1274
0.939
0.27
Table 1: Quantitative Comparison . We evaluate 3D consistency using Chamfer Distance (CD.), masked SSIM,LPIPS on unedited regions, overall 3D quality with FID and FVD, and condition alignment with DINO-I and CLIP-T. Our method consistently achieves superior structure consistency, semantic alignment with the target edited image, and generation fidelity.
Method
Text Alignment ↑
Overall 3D Quality ↑
TRELLIS
63.0%
61.5%
Hunyuan
81.0%
81.0%
Ours
91.0%
88.5%
Table 2: User study . We omit results for Tailor3D and Vox-E from the table for clarity, given the strong user preference for Hunyuan and BVE. As evidenced by the table, our method was preferred by participants.
Figure 7: Qualitative comparison . Our method achieves superior editing performance with faithful instruction-semantic alignment and remarkable original structure consistency across multi-view images. Notably, the quality of our edited 3D models is comparable to those generated by TRELLIS using editing images.
Figure 8: Ablation study on Structure/Structured Latent Editing
Figure 9: Ablation study on Mask-Enhanced Loss
Appendix figures & tables10 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 10: More examples of generative data from text-guided local editing in our proposed Edit3D-Verse dataset.
Figure 11: More examples of generative data from text-guided global editing in our proposed Edit3D-Verse dataset.
Figure 12: Editing instruction prompts
Figure 13: Image generation prompts
Figure 14: Distribution of aesthetic scores in different action types.
Figure 15: Image examples from Edit3D-Verse with their corresponding overall scores
Figure 16: 3D asset examples from Edit3D-Verse with their corresponding overall scores
Figure 17: Qualitative comparison with state-of-the-art methods. The first row displays the single input image used for inference. The other rows show the results from baselines and our method. Note that Trellis, VoxHammer, and Ours display renders of the generated 3D assets, whereas Hunyuan shows 2D image editing results. Our method achieves superior 3D consistency and fidelity to the input instruction compared to both 2D-lifting (Trellis) and 2D-editing (Hunyuan) approaches.
Figure 18: More results generated by with AI Prompts
Figure 19: More results generated by with AI Prompts
Text-guided 3D editing aims to modify existing 3D assets using natural-language instructions. Current methods struggle to jointly understand complex prompts, automatically localize edits in 3D, and preserve unedited content. We introduce Vinedresser3D, an agentic framework for high-quality text-guided 3D editing that operates directly in the latent space of a native 3D generative model. Given a 3D asset and an editing prompt, Vinedresser3D uses a multimodal large language model to infer rich descriptions of the original asset, identify the edit region and edit type (addition, modification, deletion), and generate decomposed structural and appearance-level text guidance. The agent then selects an informative view and applies an image editing model to obtain visual guidance. Finally, an inversion-based rectified-flow inpainting pipeline with an interleaved sampling module performs editing in the 3D latent space, enforcing prompt alignment while maintaining 3D coherence and unedited regions. Experiments on diverse 3D edits demonstrate that Vinedresser3D outperforms prior baselines in both automatic metrics and human preference studies, while enabling precise, coherent, and mask-free 3D editing.
Yankuan Chi, Xiang Li, Zixuan Huang +1
The Hong Kong University of Science and Technology · University of Illinois Urbana-Champaign
Local editing of 3D objects remains a long-standing challenge. When interacting with 3D content, humans naturally tend to specify a coarse region of interest for modification rather than defining precise editing boundaries. However, previous methods rely on fully edited 2D images, precise 3D masks, or redundant pipelines, which present a gap. To bridge this gap, we propose EditVerse3D, a novel 3D editing framework that enables high-quality object editing under such coarse guidance. Our approach takes as input a 3D object to be edited, a coarse 3D bounding box indicating the target region, and a reference 2D image describing the desired modification. It produces a coherent, high-fidelity edited 3D object. To facilitate this editing, we introduce a novel region-aware adaptive loss that emphasizes hard-to-learn regions and balances the objective between target and preserved areas. Complementing our loss function, we enhance model robustness and generalization through targeted data augmentations, such as training with scaled 3D masks and filtering out unrealistic editing pairs. We construct a large-scale 3D editing dataset derived from parts information. Extensive experiments demonstrate that EditVerse3D achieves superior visual quality and quantitative performance compared to existing 3D editing approaches. Please visit our project page at https://editverse3d.github.io.
Youtan Yin, Yanning Zhou, Jiacheng Wei +6
College of Computing and Data Science, Nanyang Technological University · Tencent AIPD
Controllable local editing of 3D assets requires precise target localization and appropriate visual guidance. However, existing methods lack a simple yet accurate way to obtain 3D masks and struggle to achieve the desired edit while faithfully preserving the structure and appearance of non-target regions. To address these challenges, we present EditFlow3D, a training-free framework for local 3D editing. Given a source asset and an edit instruction, a VLM-driven workflow interprets the editing intent and automatically constructs a visual guidance image and a refined 3D editing mask, enabling localized editing in the native representation space of a pretrained 3D generative model. Specifically, mask-guided differential flow focuses the edit on the target region, while step-wise trajectory preservation maintains consistency between non-target regions and the source asset without directly replacing intermediate features. Since the existing Edit3D-Bench covers only a limited range of local editing categories, we further introduce EditFlow-Bench as a complementary benchmark encompassing a broader variety of structural and appearance edits, and evaluate EditFlow3D on both benchmarks. Quantitative results, qualitative comparisons, and a user study demonstrate that EditFlow3D achieves more accurate target-region editing and better preserves non-target regions than existing 3D editing methods.
Rui Nie, Chuang Wang, Haitao Zhou +4
Beihang University · Bambu Lab · Shanghai Jiao Tong University