3D texture generation is receiving increasing attention, as it enables the creation of realistic and aesthetic texture materials for untextured 3D meshes. However, existing 3D texture generation methods are limited to producing only a few types of non-emissive PBR materials (e.g., albedo, metallic maps and roughness maps), making them difficult to replicate highly popular styles, such as cyberpunk, failing to achieve effects like realistic LED emissions. To address this limitation, we propose a novel task, emission texture generation, which enables the synthesized 3D objects to faithfully reproduce the emission materials from input reference images. Our key contributions include: first, We construct the Objaverse-Emission dataset, the first dataset that contains 40k 3D assets with high-quality emission materials. Second, we propose EmissionGen, a novel baseline for the emission texture generation task. Third, we define detailed evaluation metrics for the emission texture generation task. Our results demonstrate significant potential for future industrial applications. Dataset will be available at https://github.com/yx345kw/EmissionGen.
Figures & tables
Figure 1. Gallery of our generated emission texture 3D assets.
Figure 2. Overview of the Objaverse-Emission dataset construction pipeline. Starting from Objaverse 1.0 and Objaverse-XL, we perform (1) coarse asset filtering, (2) emission-material verification, (3) multi-strength and multi-illumination rendering, (4) luminous-area–based emission strength selection, and (5) VLM-based single-object screening. The resulting Objaverse-Emission dataset contains 40K high-quality 3D assets with reliable emission materials.
Figure 3. Dataset statistics of Objaverse-Emission: (a) word cloud of metadata tags; (b) categories distribution.
Figure 4. Model architecture of EmissionGen. Following Hunyuan3D-2.1 Paint, we use a dual-branch architecture. The input to the reference branch and emission strength branch is a reference image. The input to the generation branch are concatenation of latent noise and multi-view normal maps and position maps of corresponding 3D mesh. The output of generation branch is multi-view albedo maps and emission maps. In each generation block, there are reference attention and multi-view attention block. For the alignment of albedo maps and emission maps, the reference attention of albedo branch is shared with emission branch.
Figure 5. Glare disturbance from emission materials. The left image shows the reference rendering, while the middle and right images depict the incorrectly generated albedo and emission maps. The strong glow emitted by the sword illuminates the adjacent lap region of the warrior, and this glare is mistakenly baked into both the albedo and emission maps during generation.
Figure 6. Multi-view renderings of a 3D asset under varying emission strengths and environmental illuminations. During emission-strength disentanglement training, reference images are randomly sampled across different viewpoints, lighting conditions, and emission intensities to encourage illumination-invariant material learning.
Table 1. Quantitative evaluation on texture realism and diversity.
Figure 7. Qualitative comparison of different models. The left most column are reference images. On the right are multi-view images rendered from the complete synthesized textured 3D meshes generated by different models. The results show that EmissionGen achieves superior results compared to other methods. Please zoom in for better visualization of the details.
Method
Dice coefficient ↑
EmissionGen
0.98856
MAGE-Emission
0.98721
MVPainter-Emission
0.98749
Table 2. Emission map accuracy.
Figure 8. With ESDT (Emission Strength Disentanglement Training), the glare caused by emission materials will not be baked into albedo maps, and the generated emission maps will be correct.
Method
PSNR ↑
SSIM ↑
LPIPS ↓
CLIP-FID ↓
FID ↓
CLIP-I ↑
w/ seg loss
15.36
0.912
0.086
23.491
134.3
0.824
w/o seg loss
14.43
0.891
0.088
24.245
139.1
0.810
Table 3. Emission-channel segmentation loss ablation study.
Single image 3D generation aims to dress a given 3D mesh with relightable materials while faithfully preserving the visual details of a reference image. Existing methods either introduce artifacts when fusing inconsistent multi-view predictions or struggle to preserve fine details through implicit image-feature conditioning. To resolve these challenges, we introduce SUPRA-Tex, built on the principle of decoupling visible-material estimation, surface alignment, and observation-guided 3D completion. This division of labor allows the 3D generator, trained on limited 3D data, to focus its capacity on preserving visible evidence and completing unobserved surfaces. Specifically, we first train an MLLM-based image editing model to predict lighting-separated PBR maps in the reference view. We then project these maps onto the visible surface and encode them with the pretrained material VAE encoder into native observation latents that share the generation target's latent space. An interleaved flow transformer holds these latents as fixed, read-only anchors while alternating unobserved-region and full-surface updates to generate complete material latents. This design achieves super-alignment with the reference, delivering What You See Is What You Get fidelity on visible regions and coherent completion on unobserved surfaces. Trained solely on public datasets, SUPRA-Tex achieves SOTA performance in both geometry-conditioned texturing and end-to-end asset generation, matching or even exceeding leading commercial systems on in-the-wild images. Project page: https://supra-3d.github.io/supratex/
Chen Yang, Jiemin Fang, Fanpeng Meng +5
Huawei Inc. · Shanghai Jiao Tong University · Huazhong University of Science and Technology
Recent 3D generative models can synthesize high-quality geometry but often struggle to reproduce intricate textures from reference images, largely due to the scarcity of large-scale 3D training data with rich surface appearance. In contrast, visual generative models are trained on datasets several orders of magnitude larger and excel at modeling complex visual patterns. Motivated by this gap, we introduce Ink3D, a framework that bridges 3D generation with large-scale video generative models to synthesize extremely complex textures. Ink3D first reconstructs a white-mesh geometry using an off-the-shelf 3D generation model. It then employs OrbitPainter, a conditional video generative model, to produce dense orbit-scan videos capturing object appearance across viewpoints. To convert these views into coherent textures, we introduce TextureOptimizer, a neural baking module that integrates dense multi-view observations while mitigating geometry inconsistencies arising from video generation. By decoupling geometry and texture synthesis and leveraging large-scale pretrained video priors, Ink3D enables significantly richer and more faithful texture generation than prior approaches.
Yue Han, Chong Li, Zhening Liu +5
ZGCA & ZGCI · Zhejiang University · Microsoft Research +1
We present a method for generating physically-based materials for 3D shapes based on a video diffusion transformer architecture. Our method is conditioned on input geometry and a text description, and jointly models multiple material properties (base color, roughness, metallicity, height map) to form physically plausible materials. We further introduce a custom variational auto-encoder which encodes multiple material modalities into a compact latent space, which enables joint generation of multiple modalities without increasing the number of tokens. Our pipeline generates high-quality materials for 3D shapes given a text prompt, compatible with common content creation tools.