“Shap·E” is a novel conditional generative model developed by OpenAI for creating 3D assets. It outperforms similar models by generating parameters of implicit functions that can be rendered as textured meshes and neural radiance fields. Trained on a large dataset of 3D and text data, Shap·E is capable of generating complex and diverse 3D assets quickly and efficiently. Compared to Point·E, another 3D generative model, Shap·E achieves faster convergence and equal or better quality results. By providing a more flexible output representation, this model opens up a myriad of possibilities for downstream 3D applications.

 

Publication date: May 3, 2023
Project Page: https://github.com/openai/shap-e
Paper: https://arxiv.org/pdf/2305.02463.pdf