Documentation · 中文文档 · Contributing
WorldFoundry is an open-source framework for world-model inference and evaluation. It brings video generation, interactive worlds, 3D/4D reconstruction, and embodied action into a shared workflow.
- [2026-07-12] 🔥 WorldFoundry reached 100+ stars on its very first day! Thanks to the community for the incredible support and encouragement. More exciting updates are coming!
- [2026-07-11] 🎉 WorldFoundry is officially open-sourced. We welcome ⭐ stars, bug reports, feature requests, and pull requests from the community!
Join the WorldFoundry Community on Slack, Discord, or WeChat for discussions, announcements, technical support, and the latest project updates.
WeChat Community
| Group 1 — Full | Group 2 — Full | Group 3 — Open |
 |
 |
 |
Groups 1 and 2 are full. Scan the Group 3 QR code to join.
QR codes are updated periodically if they expire.
- Model and benchmark catalogs — discover models, checkpoint sources, runtime requirements, and benchmark protocols.
- Unified inference — run in-tree pipelines through the CLI, terminal UI, or browser-based Studio.
- Artifact inspection — review job status, generated media, and geometry through Studio and its visualizers.
- Benchmark evaluation — run model × benchmark cells and inspect normalized metrics, readiness, and eligibility scorecards.
- Platform: Linux with an NVIDIA CUDA-capable GPU.
- Driver: CUDA 12.8+ support for the default unified environment.
- GPU architecture / VRAM: model-specific; check the selected model's page.
- Tools: Git and Conda.
Inference, Studio, evaluation, and development use the same unified CUDA environment. Model-specific environments for ABI or simulator requirements are documented in the environment guide.
Clone the repository, create the GPU environment, and launch Studio:
GIT_LFS_SKIP_SMUDGE=1 git clone https://github.com/OpenEnvision/WorldFoundry.git
cd WorldFoundry
bash scripts/setup/bootstrap_worldfoundry.sh
source tmp/worldfoundry_unified_env.sh
conda activate "${WORLDFOUNDRY_UNIFIED_ENV_PREFIX}"
bash scripts/workspace/run_workspace.sh --host 127.0.0.1 --port 7870Open http://127.0.0.1:7870/ to configure inference jobs and inspect their outputs.
On a remote GPU machine, forward port 7870; additional viewer ports are listed in the
Studio guide.
Checkpoints, datasets, API keys, and generated artifacts stay outside git. Before running a model, prepare its assets using the model page and local asset guide.
Prefer the terminal? Run worldfoundry-eval tui in the same environment to build a runnable command.
The full Quickstart walks through assets,
inference, and evaluation.
Browse the documentation in English or 中文.
| Task | Guide |
|---|---|
| Choose a model or benchmark | Models · Benchmark Hub |
| Prepare environments and assets | Environments · Local assets |
| Run inference from the terminal | Inference · CLI |
| Use the browser workspace | Studio |
| Tune inference | Inference optimization |
| Score generated outputs | Evaluation |
| Validate changes and understand internals | Validation · Architecture |
See CONTRIBUTING.md for GPU development setup, coding conventions, tests, DCO sign-off, and the PR workflow.
Repository Agent Skills provide opt-in workflows for model and benchmark integration, inference optimization, numerical validation, and maintenance.
Use these checks before opening a PR or marking a model/benchmark ready:
Community standards and private vulnerability reporting are covered by the Code of Conduct and Security Policy.
If you use WorldFoundry in research, cite this repository and the upstream methods, checkpoints, datasets, and benchmarks that your work depends on.
Contributions follow the Code of Conduct. Use the Security Policy for private vulnerability reports.
WorldFoundry-authored code is licensed under the Apache License 2.0, except where a file or component carries a different license notice. Vendored and adapted upstream components remain subject to their original licenses, collected in THIRD-PARTY-NOTICES.
WorldFoundry builds on upstream world-model, video-generation, reconstruction, and embodied-action projects. See the model pages and upstream license guide for attribution.
We also thank:
- FastVideo — a unified inference and post-training framework for accelerated video generation
- OpenWorldLib — a unified codebase for advanced world models
- VLA Evaluation Harness — one framework to evaluate VLA models on robot simulation benchmarks