Step1X-3D: Towards High-Fidelity and Controllable Generation of Textured 3D Assets
demo
Step1X-3D demonstrates the capability to generate 3D assets with high-fidelity geometry and versatile texture maps, while maintaining exceptional alignment between surface geometry and texture mapping. From left to right, we sequentially present: the base geometry (untextured), followed by cartoon-style, sketch-style, and photorealistic 3D asset generation results.
🔥🔥🔥 Latest News!!
May 13, 2025: 👋 Step1X-3D online demo is available on huggingface-enjoy yourself with generated 3D assets! Huggingface web live
May 13, 2025: 👋 We release the 800K uids of high quality 3D assets (excluding self-collected assets) obtained with our rigorous data curation pipeline for both training 3D geometry and synthesis. Huggingface dataset
May 13, 2025: 👋 We have also released the training code of both Step1X-3D geometry generation and texture synthesis.
May 13, 2025: 👋 We have released the inference code and model weights of Step1X-3D geometry and Step1X-3D texture.
May 13, 2025: 👋 We have released Step1X-3D technical report as open source.
📑 Open-source Plan
Technical report
Inference code & model weights
Training code
Uid of high quality 3D assets
Online demo (gradio deployed on huggingface)
More controllable models, like conditioned with multi-view, bounding-box and skeleton
ComfyUI
1. Introduction
While generative artificial intelligence has advanced significantly across text, image, audio, and video domains, 3D generation remains comparatively underdeveloped due to fundamental challenges such as data scarcity, algorithmic limitations, and ecosystem fragmentation.
To this end, we present Step1X-3D, an open framework addressing these challenges through:
(1) a rigorous data curation pipeline processing >5M assets to create a 2M high-quality dataset with standardized geometric and textural properties;
(2) a two-stage 3D-native architecture combining a hybrid VAE-DiT geometry generator
with an SD-XL-based texture synthesis module; and (3) the full open-source release of models, training code, and adaptation modules. For geometry generation, the hybrid VAE-DiT component produces watertight TSDF representations by employing perceiver-based latent encoding with sharp edge sampling for detail preservation. The SD-XL-based texture synthesis module then ensures cross-view consistency through geometric conditioning and latent-space synchronization.
Benchmark results demonstrate state-of-the-art performance that exceeds existing open-source methods, while also achieving competitive quality with proprietary solutions.
Notebly, the framework uniquely bridges 2D and 3D generation paradigms by supporting direct transfer of 2D control techniques~(e.g., LoRA) to 3D synthesis.
By simultaneously advancing data quality, algorithmic fidelity, and reproducibility, Step1X-3D aims to establish new standards for open research in controllable 3D asset generation.
Given the above high quality 3D assets, you can follow methods from Dora to preprocess data for VAE and 3D DiT training, and from MV-Adapter for ig2mv training.
4. Dependencies and Installation
The dependencies configured according to the following instructions provide an environment equipped for both training and inference
4.1 Clone the repo
git clone https://github.com/stepfun-ai/Step1X-3D.git
cd Step1X-3D
You can select a configuration file for training and modify the scripts to support multi-GPU training or more training setting.
6.1 Train variational autoencoder
# example of VAE config in path: Step1X-3D/configs/train-geometry-autoencoder
CUDA_VISIBLE_DEVICES=0 python train.py --config $config --train --gpu 0
6.2 Train 3D native diffusion model from scratch
# example of 3D diffusion config in path: Step1X-3D/configs/train-geometry-diffusiontrain-geometry-autoencoder
CUDA_VISIBLE_DEVICES=0 python train.py --config $config --train --gpu 0
6.3 Train 3D native diffusion model with LoRA finetuning
# example of 3D ig2mv config in Path: Step1X-3D/configs/
train-texture-ig2mv
# We adopt most training code for multi-view generation from MV-Adapter and thank for the nice work.
CUDA_VISIBLE_DEVICES=0 python train_ig2mv.py --config configs/train-texture-ig2mv/step1x3d_ig2mv_sdxl.yaml --train
Step1X-3D is licensed under the Apache License 2.0. You can find the license files in the respective github and HuggingFace repositories.
9. Citation
If you find our work helpful, please cite us
@article{li2025step1x3dhighfidelitycontrollablegeneration,
title={Step1X-3D: Towards High-Fidelity and Controllable Generation of Textured 3D Assets},
author={Weiyu Li and Xuanyang Zhang and Zheng Sun and Di Qi and Hao Li and Wei Cheng and Weiwei Cai and Shihao Wu and Jiarui Liu and Zihao Wang and Xiao Chen and Feipeng Tian and Jianxiong Pan and Zeming Li and Gang Yu and Xiangyu Zhang and Daxin Jiang and Ping Tan},
journal={arXiv preprint arxiv:2505.07747}
year={2025}
}
