Note: This Replicate deployment is maintained by firtoz, a fan of the TRELLIS project, and is not officially affiliated with Microsoft or the TRELLIS team. All rights, licenses, and intellectual property belong to Microsoft. For the original project, please visit microsoft/TRELLIS.

Note: This Replicate deployment is maintained by firtoz, a fan of the TRELLIS project, and is not officially affiliated with Microsoft or the TRELLIS team. All rights, licenses, and intellectual property belong to Microsoft. For the original project, please visit microsoft/TRELLIS. TRELLIS is a powerful 3D asset generation model that converts text or image prompts into high-quality 3D assets. This Replicate deployment focuses on the image-to-3D generation capabilities of TRELLIS.

TRELLIS is a powerful 3D asset generation model that converts text or image prompts into high-quality 3D assets. This Replicate deployment focuses on the image-to-3D generation capabilities of TRELLIS.

12/26/2024

• Release TRELLIS-500K dataset and toolkits for data preparation.

12/18/2024

• Implementation of multi-image conditioning for TRELLIS-image model. (#7). This is based on tuning-free algorithm without training a specialized model, so it may not give the best results for all input images.
• Add Gaussian export in app.py and example.py. (#40)

• Release inference code and TRELLIS-image-large model
• Release dataset and dataset toolkits
• Release TRELLIS-text model series
• Release training code

TRELLIS uses a unified Structured LATent (SLAT) representation that enables generation of different 3D output formats. The model deployed here is TRELLIS-image-large, which contains 1.2B parameters and is trained on a diverse dataset of 500K 3D objects.

TRELLIS uses a unified Structured LATent (SLAT) representation that enables generation of different 3D output formats. The model deployed here is TRELLIS-image-large, which contains 1.2B parameters and is trained on a diverse dataset of 500K 3D objects. Key features:

• Generate high-quality 3D assets from 1 to N input images
• Multiview input support for enhanced 3D reconstruction accuracy
• Multiple output formats: 3D Gaussians, Radiance Fields, and textured meshes
• Detailed shape and texture generation
• Support for various viewpoint renderings

Key features:

• Generate high-quality 3D assets from input images
• Multiple output formats: 3D Gaussians, Radiance Fields, and textured meshes
• Detailed shape and texture generation
• Support for various viewpoint renderings For more examples and to try it directly in your browser, visit the Replicate model page.

For more examples and to try it directly in your browser, visit the Replicate model page.

The model accepts:

• 1 to N input images (PNG or JPEG format) for multiview 3D reconstruction
• Optional parameters for controlling the generation process

The model accepts:

• An input image (PNG or JPEG format)
• Optional parameters for controlling the generation process

The model outputs:

• A GLB file containing the generated 3D model with textures
• Preview renders from multiple angles
• Background-removed versions of all input images
• Optional: Raw 3D Gaussians or Radiance Field representations

The model outputs:

• A GLB file containing the generated 3D model with textures
• Preview renders from multiple angles
• Optional: Raw 3D Gaussians or Radiance Field representations

import replicate

output = replicate.run(
    "fire/trellis:version",
    input={
        "seed": 0,
        "images": [
            "https://replicate.delivery/pbxt/M6rvlcKpjcTijzvLfJw8SCWQ74M1jrxowbVDT6nNTxREcvxO/ephemeros_cartoonish_character_art_cyberpunk_crocodile_white_ba_486fb649-bc68-46a0-b429-751b43734b89.png",
        ],
        "texture_size": 1024,
        "mesh_simplify": 0.95,
        "generate_color": True,
        "generate_model": False,
        "save_gaussian_ply": True,
        "randomize_seed": True,
        "generate_normal": True,
        "ss_sampling_steps": 12,
        "slat_sampling_steps": 12,
        "ss_guidance_strength": 7.5,
        "slat_guidance_strength": 3
    }
)
print(output)

If you cannot cog push in Docker Desktop, try disabling containerd.

app.py provides a simple web demo for 3D asset generation. Since this demo is based on Gradio, additional dependencies are required:

. ./setup.sh --demo

After installing the dependencies, you can run the demo with the following command:

python app.py

Then, you can access the demo at the address shown in the terminal.

The web demo is also available on Hugging Face Spaces!

We provide TRELLIS-500K, a large-scale dataset containing 500K 3D assets curated from Objaverse(XL), ABO, 3D-FUTURE, HSSD, and Toys4k, filtered based on aesthetic scores. Please refer to the dataset README for more details.

TRELLIS models and the majority of the code are licensed under the MIT License. The following submodules may have different licenses:

• diffoctreerast: We developed a CUDA-based real-time differentiable octree renderer for rendering radiance fields as part of this project. This renderer is derived from the diff-gaussian-rasterization project and is available under the LICENSE.

• Modified Flexicubes: In this project, we used a modified version of Flexicubes to support vertex attributes. This modified version is licensed under the LICENSE.

If you find this work helpful, please consider citing our paper:

@article{xiang2024structured,
    title   = {Structured 3D Latents for Scalable and Versatile 3D Generation},
    author  = {Xiang, Jianfeng and Lv, Zelong and Xu, Sicheng and Deng, Yu and Wang, Ruicheng and Zhang, Bowen and Chen, Dong and Tong, Xin and Yang, Jiaolong},
    journal = {arXiv preprint arXiv:2412.01506},
    year    = {2024}
}

• Project Page
• Paper
• GitHub Repository
• Hugging Face Demo TRELLIS is released under the MIT License. See LICENSE for details.

• Project Page
• Paper
• GitHub Repository
• Hugging Face Demo