SceneDreamer: A Powerful 3D Scene Generation Tool

SceneDreamer is a tool developed by Nanyang Technological University that generates stunning 3D scenes from 2D images.

What Makes SceneDreamer Special

Unconditional Generative Model

SceneDreamer can create large 3D landscapes from 2D images without specific conditions.

Learning Paradigm

It utilizes an efficient 3D scene representation, generative scene parameterization, and a powerful renderer.

Efficient Bird's-Eye-View Representation

SceneDreamer efficiently processes surface elevation and detailed scene semantics.

Generative Neural Hash Grid

Features are consistent and content is efficiently encoded across different scenes.

Neural Volumetric Renderer

It produces visually stunning, photorealistic images of the scenes it generates.

Pros and Cons of SceneDreamer


  • Enables unbounded 3D scene generation from 2D images.
  • Learns from in-the-wild 2D image collections without the need for 3D annotations.
  • Purposeful and detailed approach to efficient representation of 3D scenes.
  • Consistent features and well-encoded content across different scenes.
  • Impressively photorealistic and visually captivating results.


  • More research-oriented and not a commercial tool for everyday use.
  • Requires a degree of understanding in computer science and AI.


SceneDreamer is an AI-powered system that transforms 2D images into extraordinary 3D scenes. Whether you're a researcher, developer, or just curious about AI, SceneDreamer is an incredible tool to explore. The team at Nanyang Technological University has developed groundbreaking technology with endless future possibilities.