Skip to content

DiffSplat

Generate 3D Gaussian point clouds from text or one image

Visit official site

What DiffSplat is

The framework is described as a 3D technique that turns text prompts and single-view images into Gaussian point clouds, reusing large pretrained text-to-image diffusion models.

What you can do with it

  • Generate point clouds from a text prompt
  • Generate point clouds from a single-view image
  • Adjust style with ControlNet

Application scenarios

  • A rainbow fish model comes from a text prompt
  • A toy robot model comes from one image
  • ControlNet gives a robot a steampunk look

Pros & cons

✓ What we like

  • ControlNet allows style adjustment
  • Several pretrained 2D diffusion models are compatible

! What to watch out for

  • Outputs are 3D Gaussian point clouds
  • Generation relies on pretrained 2D diffusion models

FAQ

What inputs are used?

The material names text prompts and single-view images.

How fast is it?

The material states one to two seconds for generation.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review