“Poxels”: Researchers enable detailed color printing of AI-generated 3D models

Printing AI-generated 3D models is still a challenge in practice. Automated processes often only deliver simple geometries without detailed material information. Although these are suitable for decorative purposes, they rarely meet technical requirements for full-color and high-quality printing results.

A research group from Japan has now presented a new method for outputting AI-generated models directly as printable objects. The core of this technology is the introduction of so-called “Poxels”, short for “printable voxels”. These printable voxels not only contain spatial information, but also specific material and color values that are required for 3D printing. Previously existing methods for generating 3D models from text or images only provided surface textures in the RGB color space. Although this data can be displayed on the screen, it is only suitable for physical printing to a limited extent.

The scientists therefore developed a process that specifically translates RGB color values into the physical color space CMYKWCl. This color space is used specifically by polymer or binder jetting printing systems. In these printing processes, individual voxels can be controlled separately, allowing detailed and color-accurate models to be produced. Processes such as the widespread fused filament fabrication (FFF) or SLA printing are not compatible due to their technical functionality.

To test their method, the researchers used the Stratasys J850 full-color 3D printer, which can mix and print colors precisely at the voxel level. In practical tests, the models printed on the basis of “poxels” achieved a high level of color fidelity between the digital template and the physical object. The method proved to be clearly superior to existing approaches, such as “3D Gaussian splatting”, particularly in terms of precision and color reproduction.

Details can be found in the paper entitled “Poxel: Voxel Reconstruction for 3D Printing”.