GaussianDreamer

Why we included this project

Turning a text prompt into a 3D object usually takes hours with older NeRF-based methods, but GaussianDreamer gets you a Gaussian-splatting asset in about 15 minutes on a single GPU. It works by seeding the scene with a 3D diffusion prior, then letting a 2D diffusion model refine geometry and appearance, a combination worth studying if you're building your own text-to-3D pipeline. The output renders in real time and can be dropped into Unity, so it fits game and design workflows directly. The repo also covers text-to-avatar generation via SMPL initialization, and the authors point to GaussianDreamerPro as a higher-quality follow-up for animation and simulation. Expect research code with a CVPR 2024 paper behind it: you'll handle installation and tuning yourself rather than getting a turnkey product.