ИИ 3D
gaussian-splatting-lightning
A 3DGS framework with derived algorithms and a web viewer.
★ 1,1 тыс. звёздNOASSERTION
Смотреть руководство →Nerfstudio — проект с открытым исходным кодом из категории «ИИ 3D» от nerfstudio-project (nerfstudio-project/nerfstudio на GitHub). У него 12 083 звезды на GitHub и 1 681 форк, он распространяется по лицензии Apache-2.0 и последний раз обновлялся на GitHub 29 июля 2025.
Nerfstudio — the plug-and-play studio for NeRFs and Gaussian splatting: turn photos of a real scene into explorable 3D, with a modular API and viewer, Apache-2.0
Per the README: you need an NVIDIA GPU with CUDA; create the conda env (python 3.8 in the README setup), then install PyTorch with CUDA and tiny-cuda-nn.
pip install nerfstudio, then process your photos/video into a dataset with the ns-process-data tools (COLMAP pose estimation under the hood).
Train a method (Nerfacto is the default workhorse; Splatfacto for Gaussian splatting) and inspect, render and export from the built-in viewer.
Data processing, training, a real-time viewer, camera-path rendering and export in one coherent tool.
First-class methods for both families — Nerfacto for NeRF, Splatfacto for 3D Gaussian splatting.
Components are interpretable and swappable, so new methods plug in — the reason the research community builds on it.
Explore the scene as it trains, set camera paths and export renders without leaving the tool.
The README environment setup.
conda create --name nerfstudio -y python=3.8
conda activate nerfstudio
pip install --upgrade pipCUDA 11.8 build per the README, then the package itself (tiny-cuda-nn is covered in the README dependencies section).
pip install torch==2.1.2+cu118 torchvision==0.16.2+cu118 --extra-index-url https://download.pytorch.org/whl/cu118
pip install nerfstudioThe docs quickstart trains Nerfacto on the classic Blender scene, then on your own captures.
https://docs.nerf.studio/quickstart/first_nerf.htmlThe Nerfstudio project, launched by Berkeley students in the KAIR lab at BAIR and now developed with community contributors (nerfstudio-project on GitHub); Luma AI and BAIR Commons are listed sponsors.
Yes — Apache-2.0 licensed.
An NVIDIA GPU with CUDA installed — the README tested setup is CUDA 11.8 with PyTorch 2.1.2. CPU-only training is not the intended path.
Against the original gaussian-splatting research repo, Nerfstudio is the full studio — data processing, viewer, many methods, exports — rather than a single-method implementation. Against hosted capture apps (Luma, Polycam), it is free, local and export-friendly, at the cost of running the training yourself.
Follow the README: conda env, PyTorch with CUDA, tiny-cuda-nn, pip install nerfstudio — then run the docs quickstart on the Blender scene before your own photos.
A 3DGS framework with derived algorithms and a web viewer.
Gaussian splatting operators that bring 3DGS into SideFX Houdini.
Cross-vendor 3D Gaussian Splatting trainer: video to splat to mesh.
ICLR 2024] Live Photorealistic Dynamic Scene Representation and Rendering.
Real-time LiDAR-inertial-camera Gaussian Splatting SLAM.
LiDAR-augmented Gaussian Splatting plus neural SDF, in real time.
Другие проекты из категории «ИИ 3D» с наибольшим числом звёзд в каталоге LibreAI — реалистичный список альтернатив, если Nerfstudio вам не подходит.
Control Blender with natural language: an MCP plugin that lets any LLM model, light and animate in 3D
Pascal Editor: an open-source 3D architectural editor with MCP tools for humans and AI agents
Give your AI agent CAD superpowers - generate real STEP/STL models from text
Turn a reference image into a procedural, animation-ready Three.js model - as code
High-Resolution 3D Assets Generation with Large Scale Hunyuan3D Diffusion Models.
Соседние проекты из категории «ИИ 3D» в каталоге, ближайшие по числу звёзд к Nerfstudio — продолжайте изучать полку по обе стороны от него.
Shap-E — OpenAI's text-to-3D and image-to-3D model: generate 3D objects as implicit functions and render them as NeRFs or textured meshes, MIT
Modly — a desktop app that generates 3D models from images or prompts, running entirely on your own GPU.
Official repo for paper "Structured 3D Latents for Scalable and Versatile 3D Generation" (CVPR'25 Spotlight).
From Images to High-Fidelity 3D Assets with Production-Ready PBR Material
High-Resolution 3D Assets Generation with Large Scale Hunyuan3D Diffusion Models.
Tencent Hunyuan3D-1.0: A Unified Framework for Text-to-3D and Image-to-3D Generation
Бесплатно и открыто — забирайте исходный код, ставьте звезду репозиторию и начинайте создавать.