Diffusion-Guided 3D Asset Generation

Explore diffusion-guided generation of textured 3D assets through modern image and 3D modeling techniques.

Tech Used

Stable DiffusionDiffusersControlNetLoRADreamFusion (SDS)NeRF/Instant-NGPHugging Face TransformersTripoSRPyTorch3DOpen3DBlendertrimeshCUDAONNXTensorRTThree.jsWebGL/WebGPUFastAPIGradio

Problem

Creating detailed 3D assets is slow and specialized, while text- or image-guided generation can accelerate ideation and prototyping.

Approach

Used diffusion-based guidance methods such as ControlNet and LoRA for detail generation
Explored 3D representations and reconstruction approaches including NeRF-style and splatting-inspired workflows
Connected experimental components into prototype interfaces for testing end-to-end generation ideas

Results

Research-oriented pipeline for faster 3D asset ideation
Reusable experimental foundation for generative 3D workflows

Screenshots