deepseek-ai/DreamCraft3D
A hierarchical 3D generation method using bootstrapped diffusion priors to produce high-fidelity coherent 3D objects from 2D reference images.
Not currently ranked — collecting fresh signals.
star history
DreamCraft3D generates 3D content by leveraging a 2D reference image to guide geometry sculpting and texture boosting stages. It uses view-dependent diffusion models for score distillation sampling to ensure geometry consistency, and employs a bootstrapped score distillation approach where a personalized Dreambooth model is trained on augmented scene renderings to provide view-consistent texture guidance. The system alternates between optimizing the diffusion prior and the 3D scene representation for mutually reinforcing improvements.
Frequently asked
- What is deepseek-ai/DreamCraft3D?
- A hierarchical 3D generation method using bootstrapped diffusion priors to produce high-fidelity coherent 3D objects from 2D reference images.
- Is DreamCraft3D open source?
- Yes — deepseek-ai/DreamCraft3D is open source, released under the MIT license.
- What language is DreamCraft3D written in?
- deepseek-ai/DreamCraft3D is primarily written in Python.
- How popular is DreamCraft3D?
- deepseek-ai/DreamCraft3D has 3k stars on GitHub.
- Where can I find DreamCraft3D?
- deepseek-ai/DreamCraft3D is on GitHub at https://github.com/deepseek-ai/DreamCraft3D.