Make 3D stuff

These models generate 3D content like objects, scenes, and textures. They enable creating 3D assets from text prompts and images. Key capabilities include:

  • 3D shape generation: Create 3D shapes from text prompts and images.
  • Multi-view rendering: Generate consistent views of 3D objects from different angles.
  • Texturing: Generate textures for 3D assets based on text prompts.
  • Radiance fields: Represent 3D scenes as neural radiance fields for view synthesis.

Our Pick: DreamGaussian

For most people, we recommend DreamGaussian as the best all-around model for 3D content generation. It’s the most popular model in this category, indicating wide applicability and user satisfaction.

DreamGaussian runs reasonably quickly, generating results in about 2.5 minutes on average. It was released recently, so it incorporates modern techniques. Use DreamGaussian for high-quality 3D object and scene generation from text and image prompts.

Fastest Results: Zero123++

If you prioritize speed, check out Zero123++. It’s the second most popular 3D model and generates results in just 15 seconds on average - 10x faster than DreamGaussian. Zero123++ specializes in generating multi-view images of an object from a single input image.

Cutting-Edge Research: Shap-E

Shap-E is a brand new model that represents the state-of-the-art in 3D generation research. It’s already the 3rd most popular model despite being released less than a week ago. Shap-E is a bit slower than our top picks, but worth trying for the latest and greatest 3D generation capabilities.

LGM is another fast and popular choice for 3D content creation, though it’s a bit older than our top recommendations. Wonder3D has also seen significant usage, but it’s one of the slowest models, so we don’t recommend it for most users.