Shap-E: 3D generator — what it does and what you need to run it

Shap-E is an open source 3D generator from OpenAI. The tool creates 3D objects from a text description or an input image, i.e. it solves the task of converting text/image into three-dimensional geometry.

What it does

As described by the authors, Shap-E generates 3D objects conditioned on text or images. The output is three-dimensional models that can be used in graphics applications. The authors do not specify details about supported output formats, geometry types (mesh, point cloud), or maximum texture resolution in the repository.

What you need to run it

The code is written in Python and is distributed under the MIT license. You will need a local machine with sufficient GPU VRAM — the authors do not specify exact requirements, but generative models of this class typically require several gigabytes of VRAM. There is no cloud version, running on your own hardware only. The source code is available in the repository github.com/openai/shap-e. The last code update is dated June 22, 2024.

Who it is for

The tool will be useful for developers and researchers experimenting with 3D content generation who are willing to work with Python code. The open MIT license allows embedding Shap-E into your own projects and adapting it to your needs. It is not suitable for artists and designers expecting a ready-made application with a GUI — the authors do not provide a standalone build. It is also not the right tool for those who do not have a GPU with sufficient memory or who are looking for a cloud service without the need for local installation. Shap-E occupies the niche of an open source research tool. It provides an opportunity to explore approaches to 3D generation from text and images, but requires technical expertise to run and integrate.