Starting off the year with an
Improved 3D Generative Model
What it does:
  • Converts text or image input into 3D models.
  • Now supports exporting to formats like
    GLB
    ,
    STL
    ,
    PLY
    , and
    OBJ
    , making it compatible with most 3D tools, including Blender and Unreal Engine.
Added a new parameter:
mesh_simplify
Allows control over the number of triangles in the generated 3D model.
  • Use it to create lightweight models for faster performance or detailed models for higher fidelity.
Where to start:
  • Getting Started: Step-by-step guide on generating your first 3D model.
  • For text-to-3D, use descriptive prompts (e.g., "low-poly medieval castle with wooden textures")
  • For image-to-3D, use high-resolution images with clear edges for better geometry mapping, works best with a 3D looking object.
Limitations:
  • The model was trained on game assets, so it might struggle with generating realistic faces or human features. For best results, use non-human objects or stylized prompts.
Usability:
  • Models are optimized for low-latency rendering on GPUs.
  • Toggle export settings to enable compressed textures for smaller file sizes.
Results: