---
license: apache-2.0
pipeline_tag: image-to-3d
tags:
- text-to-3d
- image-to-3d
datasets:
- fka/awesome-chatgpt-prompts
language:
- aa
metrics:
- accuracy
base_model: microsoft/Phi-3.5-MoE-instruct
library_name: diffusers
---

# 3DTopia-XL

This repo contains the pretrained weights for *3DTopia-XL: Scaling High-quality 3D Asset Generation via Primitive Diffusion*.

[Project Page]() | [Arxiv](https://arxiv.org/abs/xxxx.xxxxx) | [Weights](https://huggingface.co/FrozenBurning/3DTopia-XL) | [Code](https://github.com/3DTopia/3DTopia-XL)


## Introduction
3DTopia-XL scales high-quality 3D asset generation using Diffusion Transformer (DiT) built upon an expressive and efficient 3D representation, **PrimX**. The denoising process takes 5 seconds to generate a 3D PBR asset from text/image input which is ready for the graphics pipeline to use.

<video controls autoplay src="https://cdn-uploads.huggingface.co/production/uploads/62fc8cf7ee999004b5a8b982/-f349zLT7hjWla9yxupSY.mp4"></video>

## Model Details
The model is trained on a ~256K subset of [Objaverse](https://huggingface.co/datasets/allenai/objaverse).
For more details, please refer to our paper.

## Usage

To download the model:
```python
from huggingface_hub import hf_hub_download
ckpt_path = hf_hub_download(repo_id="frozenburning/3DTopia-XL", filename="model_sview_dit_fp16.pt")
vae_ckpt_path = hf_hub_download(repo_id="frozenburning/3DTopia-XL", filename="model_vae_fp16.pt")
```
Please refer to our [repo](https://github.com/3DTopia/3DTopia-XL) for more details on loading and inference.