Take Text to 3D Mesh with OpenAI’s New AI-Driven, Open-Source Release

OpenAI’s new release enables users to turn images or text into mesh-like 3D point clouds “in only one-to-two minutes on a single GPU.”

The AI-driven Point-E, which is available free on GitHub and fully open source, leverages “diffusion models” to create these 3D point clouds much more quickly than with advanced 3D model generators. The point clouds can then be converted to a mesh with a “regression-based model.”

An accompanying paper offers a deep-dive exploration of the technology and states that the “approach can serve as a starting point for further work in the field of text-to-3D synthesis.”

Overall, it seems Point-E has strong potential for 3D printing – although the 3D models generated are rather low resolution, and don’t show fine-grained textures or shapes. This means that point clouds created with Point-E are not currently 3D printable.

Still, with the ability to download and try for yourself, there’s not much downside to testing the waters for what tech like Point-E could lead to in the near future.

You’ve read that; now read these:

• What Does An AI Think of 3D Printing?
• Arc Overhangs Let You 3D Print on Top of ‘Thin Air’
• Mod Your Prusa for 4-Axis 3D Printing: A Closer Look