Zero123
Single-image to 3D: generate novel views of any object from one photo.
About
Zero123 (Zero-1-to-3) is a view-conditioned image generation model that produces novel views of an object from a single input image, demonstrating 3D understanding of appearance from various angles. The original research version came from Columbia and UCSD researchers; Stability AI released its in-house trained Stable Zero123 in December 2023.
Stable Zero123 is based on Stable Diffusion 1.5, uses improved training data filtered from Objaverse, and adds elevation conditioning for higher quality predictions. The weights were released for non-commercial and research use, with a commercial variant available. It became a foundational building block for single-image 3D pipelines and was later succeeded by Stable Video 3D.
What it can do
- Novel view synthesis from a single image
- Elevation-conditioned view prediction
- Score-distillation-based 3D mesh reconstruction
- Foundation for image-to-3D pipelines
Setup
Download the weights from Stability AI or Hugging Face and run inference with a diffusion pipeline.
Prerequisites
Python, PyTorch, and a GPU; 24GB VRAM recommended for full 3D reconstruction.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 103; Trending score 72/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.