InteX

This repository contains the official implementation for InteX: Interactive Text-to-Texture Synthesis via Unified Depth-aware Inpainting.

Project Page | Arxiv

Install

pip install -r requirements.txt

# nvdiffrast
pip install git+https://github.com/NVlabs/nvdiffrast/

Usage

### generate texture for a mesh based on a prompt (command line), output will be saved to ./logs
# please check ./configs/revani.yaml for each parameter's meaning
python main.py --config configs/revani.yaml mesh=data/dragon.glb prompt="a red pet dragon with fire patterns" save_path=dragon_fire.glb text_dir=True

### open local GUI for interactive editing
python main.py --config configs/revani.yaml mesh=data/dragon.glb prompt="a red pet dragon with fire patterns" save_path=dragon_fire.glb text_dir=True gui=True

### simple web GUI with gradio
python app.py

Please check ./scripts for more examples.

Acknowledgement

This work is built on many amazing research works and open-source projects, thanks a lot to all the authors for sharing!

• nvdiffrast
• dearpygui

Citation

@article{tang2024intex,
  title={InTeX: Interactive Text-to-Texture Synthesis via Unified Depth-aware Inpainting},
  author={Tang, Jiaxiang and Lu, Ruijie and Chen, Xiaokang and Wen, Xiang and Zeng, Gang and Liu, Ziwei},
  journal={arXiv preprint arXiv:2403.11878},
  year={2024}
}