Hitem3D releases improved version 2.0 of its AI 3D model generator

The AI company Hitem3D has released version 2.0 of its 3D generation service. The platform is aimed at users who want to create printable or processable 3D models from existing images. Unlike some competing systems, Hitem3D does not rely on text-to-3D, but on image-based methods. Users upload a single image or provide multiple views of an object, from which the system reconstructs a complete 3D model.

The multi-view approach is said to be particularly advantageous for complex geometries. Multiple viewing angles reduce the need to estimate hidden areas. This allows shapes and proportions to be derived more consistently. In addition, Hitem3D continues to offer a model specialized for portraits, which is designed for human head geometries. Following versions 1.0 and 1.5, version 2.0 now marks a significant technical upgrade.

At the heart of the new version is integrated texture generation. Previous workflows often first create a mesh and then project textures onto it. According to Hitem3D, this approach leads to problems such as distorted textures, visible seams, or missing details in hidden areas in high-resolution applications such as 3D printing or manufacturing. Version 2.0 therefore relies on structure-dependent texture generation, in which textures are created during geometry reconstruction.

Another new feature is lighting analysis. Light and shadow provide important clues about surface details, but can also mislead AI systems. Hitem3D therefore integrates a step to remove lighting influences before the actual 3D generation begins. This should lead to more stable results, especially with unplanned or simply taken photos.

In addition, the company has further refined the generation of human heads. The focus was on fine details and clean transitions between skin and hair areas. Overall, version 2.0 aims to reduce manual post-processing effort and increase the quality of image-based 3D models for more demanding applications.