Stable Zero123
| Property | Value |
|---|---|
| Developer | Stability AI |
| Model Type | Latent Diffusion Model |
| Base Model | lambdalabs/sd-image-variations-diffusers |
| License | Stability AI Non-Commercial Research Community License |
| Training Data | Objaverse Dataset |
What is stable-zero123?
Stable Zero123 is an advanced text-to-3D generation model developed by Stability AI that represents a significant improvement over the original Zero123 model. It specializes in view-conditioned image generation, allowing users to create high-quality 3D models from single images using Score Distillation Sampling (SDS).
Implementation Details
The model was trained on a single node with 8 A100 80GB GPUs using a modified version of the original zero123 repository. It leverages the Objaverse dataset with enhanced rendering methods to achieve superior results compared to both Zero123 and Zero123-XL.
- Improved data rendering pipeline
- Enhanced model conditioning strategies
- Integration with threestudio for 3D mesh generation
- Two versions available: Standard (non-commercial) and Commercial (Stable Zero123C)
Core Capabilities
- Single-image to 3D object generation
- Text-to-3D generation via SDXL integration
- View-conditioned image synthesis
- High-quality 3D model creation using Score Distillation Sampling
Frequently Asked Questions
Q: What makes this model unique?
Stable Zero123 stands out due to its improved rendering capabilities and conditioning strategies, offering better performance than its predecessors while maintaining the ability to generate 3D content from single images.
Q: What are the recommended use cases?
The model is ideal for researchers and developers working on 3D object generation, particularly when starting from single images or text descriptions. It's especially useful in applications requiring view-conditioned image generation and 3D model creation.
