Texture map production is an important part of 3D modeling and determines the rendering quality. Recently, diffusion-based methods have opened a new way for texture generation. However, restricted control flexibility and limited prompt modalities may prevent creators from producing desired results. Furthermore, inconsistencies between generated multi-view images often lead to poor texture generation quality. To address these issues, we introduce FlexPainter, a novel texture generation pipeline that enables flexible multi-modal conditional guidance and achieves highly consistent texture generation. A shared conditional embedding space is constructed to perform flexible aggregation between different input modalities. Utilizing such embedding space, we present an image-based CFG method to decompose structural and style information, achieving reference image-based stylization. Leveraging the 3D knowledge within the image diffusion prior, we first generate multi-view images simultaneously using a grid representation to enhance global understanding. Meanwhile, we propose a view synchronization and adaptive weighting module during diffusion sampling to further ensure local consistency. Finally, a 3D-aware texture completion model combined with a texture enhancement model is used to generate seamless, high-resolution texture maps. Comprehensive experiments demonstrate that our framework significantly outperforms state-of-the-art methods in both flexibility and generation quality.
Figures & tables
Figure 2: Example of modality aggregation and semantic manipulation by controlling the image embedding strength α . Each case’s first and second rows show text-to-image and image-to-image interpolation, respectively.
Figure 3: Pipeline of our method. We first generate multi-view images using the conditional input from the user. The top two cases show the generation of using text-only or image-only conditions. Leveraging the linear operation in Equ.( 4 ), we can also perform text-guided image refinement (shown in green) and stylization using reference image (shown in blue). The right side shows our view-synchronization and weighting module. Consistent multi-view images can be generated by reprojection and weighting during each sampling step.
Figure 4: Qualitative results on text-to-texture generation.
Figure 5: Qualitative results on image-to-texture generation.
FID ↓
KID ↓
User Preference
Runtime
TEXTure
78.269
107.062
4.4%
251
Text2Tex
89.008
94.423
11.3%
996
SyncMVD
73.385
41.737
22.5%
113
Paint3D
77.801
123.605
17.8%
204
TEXGen
72.900
61.729
15.7%
52
Ours
71.621
58.465
28.3%
153
Table 1: Quantitative evaluation result of text-to-texture generation. The best results are in bold, while the second-best ones are underlined. KID metric is multiplied by 10−4 and the unit of runtime is seconds.
FID ↓
KID ↓
User Preference
Runtime
Paint3D
83.977
267.132
28.6%
317
Ours
59.492
62.089
71.4%
157
Table 2: Quantitative evaluation result of image-to-texture generation. The best results are indicated in bold. KID metric is multiplied by 10−4 and the unit of runtime is seconds.
Figure 6: Qualitative results of our ablation study. Our full model achieves consistent and high-quality generation results, while the ablated methods suffer from ghosting artifacts and degraded quality.
Figure 7: Comparison on the stylization strategy.
Figure 8: Analysis of the intermediate results from our WeighterNet. Weights and colors from different views are spliced horizontally. The first three pictures shows the normal, condition image and the render of the final generated results. Below shows the traditional cosine weights. We can see that it lacks generalizability and fails to capture misalignments, causing blurry output. Differently, images of the other three corners shows the color UV map and the corresponding adaptive weights of different diffusion time steps. From these images we can tell that the adaptive weights can detect misalignments and solve in a learning-based, generalizablr manner.
Figure 9: Our Applications includes tasks of text-to-texture, image-to-texture, text-guided image refinement, and reference image-based stylization. More cases of our applications including the cases shown in the teaser can be seen in the appendix.
w/o view-sync
w/o WeighterNet
w/o image CFG
our full model
FID ↓
80.635
76.688
79.041
71.621
KID ↓
77.883
59.674
65.482
58.465
Table 3: Quantitative ablation studies on our view synchronization, WeighterNet, and image-based CFG. The best results are indicated in bold. KID multiplied by 10−4 .
Appendix figures & tables9 assets
Supplementary material from the paper’s appendix.
Appendix
Figure S1 : Detailed pipeline of the texture completion and enhancement module. We use a 3D-aware texture completion module containing UV blocks and point blocks to complete the partial texture. Then, we use a texture enhancement module to generate the super-resolution results.
Figure S2 : More cases of our image-to-texture generation.
Figure S3 : More cases of our text-to-texture generation.
Figure S4 : More cases of our text-to-texture generation.
Figure S5 : More cases of our imgae-to-texture generation.
Although recent advances have improved the quality of 3D texture generation, existing methods still struggle with incomplete texture coverage, cross-view inconsistency, and misalignment between geometry and texture. To address these limitations, we propose Hitem3D 2.0, a multi-view guided native 3D texture generation framework that enhances texture quality through the integration of 2D multi-view generation priors and native 3D texture representations. Hitem3D 2.0 comprises two key components: a multi-view synthesis framework and a native 3D texture generation model. The multi-view generation is built upon a pre-trained image editing backbone and incorporates plug-and-play modules that explicitly promote geometric alignment, cross-view consistency, and illumination uniformity, thereby enabling the synthesis of high-fidelity multi-view images. Conditioned on the generated views and 3D geometry, the native 3D texture generation model projects multi-view textures onto 3D surfaces while plausibly completing textures in unseen regions. Through the integration of multi-view consistency constraints with native 3D texture modeling, Hitem3D 2.0 significantly improves texture completeness, cross-view coherence, and geometric alignment. Experimental results demonstrate that Hitem3D 2.0 outperforms existing methods in terms of texture detail, fidelity, consistency, coherence, and alignment.
High-quality texture generation is essential for creating realistic and production-ready 3D assets. Recent multi-view diffusion methods have shown promising results for image-guided 3D texturing, but they are typically constrained to low operating resolutions such as 512 or 768, making it difficult to preserve high-frequency details from high-resolution reference images. Scaling this paradigm to 2048 resolution is computationally prohibitive, as the unified multi-view sequence exceeds 212K tokens and incurs excessive memory and latency. In this paper, we present UltraTex, an efficient end-to-end framework for high-resolution multi-view diffusion-based 3D texturing. Our key observation is that object-centric multi-view renderings contain two major sources of redundancy: background-induced sequence redundancy and sparse token interactions within the foreground. To address them, we introduce Background Token Dropping, which removes background tokens before the DiT backbone, and Block-Sparse Attention, which reduces attention computation over the retained foreground sequence. To enable efficient foreground-only inference while avoiding reconstruction artifacts, we further design Foreground-Aware VAE Decoding to ensure the quality of the final high-resolution views. To satisfy the demanding data requirements of 2K-resolution multi-view diffusion training, we construct G-buffer TexVerse, a large-scale, ultra-high-resolution multi-view rendering dataset covering over 268,000 3D assets. Extensive experiments show that UltraTex generates visually faithful textures with rich fine-grained details, while substantially improving efficiency, achieving 20.6×--91.1× training speedup and 22.3×--74.6× end-to-end inference speedup over the baseline on common samples in our dataset. Code and data is at https://yiboz2001.github.io/UltraTex.
Yibo Zhang, Ze Yuan, Nan Cao +4
Jilin University, Changchun, China · Shanghai Innovation Institute, Shanghai, China · The University of Hong Kong, Hong Kong, China +3
Recent 3D generation models can produce accurate geometries while still struggling to reconstruct detailed textures. We propose a diffusion-based native 3D material generation model TaoTex, which faithfully recovers intricate textures through tailored strategies and improvements. First, we develop a data construction agent to create high-frequency textured 3D assets to bridge the data gap in public datasets. Training with these data significantly enhances the ability of TaoTex to recover challenging details such as text and patterns. Second, we design a multi-level feature fusion (MLFF) module to adaptively integrate local and global features of the conditional input, providing more complete texture cues for the diffusion model and thereby enhancing reconstruction fidelity. To alleviate VAE reconstruction errors, we adopt a latent-to-pixel space loss transition, further improving the pixel-level details and generation quality. Finally, we scale TaoTex to multi-view inputs by incorporating learnable viewpoint embeddings, achieving accurate and consistent material reconstruction across views. Extensive experiments demonstrate that our method significantly outperforms existing approaches in preserving texture details in both single- and multi-view settings.