[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-93141":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":16,"forks30d":16,"starsTrendScore":16,"compositeScore":19,"rankGlobal":10,"rankLanguage":10,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":21,"hasPages":23,"topics":24,"createdAt":10,"pushedAt":10,"updatedAt":35,"readmeContent":36,"aiSummary":37,"trendingCount":16,"starSnapshotCount":16,"syncStatus":38,"lastSyncTime":39,"discoverSource":40},93141,"WorldFoundry","OpenEnvision\u002FWorldFoundry","OpenEnvision","Unified World Model Inference & Evaluation Infrastructure","https:\u002F\u002Fopenenvision.github.io\u002FWorldFoundry\u002F",null,"Python",200,12,118,3,0,5,53,51.14,"Apache License 2.0",false,"main",true,[25,26,27,28,29,30,31,32,33,34],"3d","4d","benchmark","evaluation","inference","infrastructure","video-generation","vla","wam","world-model","2026-07-22 04:02:08","# WorldFoundry\n\n[![Python](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.10%20--%203.13-blue)](pyproject.toml)\n[![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-Apache--2.0-green)](LICENSE)\n[![CLI](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCLI-worldfoundry--eval-purple)](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Freference\u002Fcli.mdx)\n[![Docs](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdocs-Fumadocs-orange)](docs\u002Ffumadocs)\n\nWorldFoundry is an open-source infrastructure for world models: a shared stack for in-tree runners, local asset staging, inference (TUI \u002F CLI \u002F Studio), and benchmark evaluation across video generation, 3D\u002F4D representation, embodied action, and interactive worlds.\n\n> ⚠️ This repository is still under active development. We will keep updating it regularly. Feel free to open an issue if you encounter any problem.\n\nDay-one workflow:\n\n1. **Environment + assets** — bootstrap conda, stage checkpoints and datasets outside git.\n2. **Inference** — generate and inspect artifacts via TUI, CLI, scripts, or Studio.\n3. **Evaluation** — score only after artifacts match the benchmark layout; use scorecards for readiness claims.\n\n## 🤝 Community\n\nJoin the **WorldFoundry Community** [Slack \u002F Wechat] for discussions, announcements, technical support, and the latest project updates.\n\n\u003Cp align=\"center\">\n\n\u003Ca href=\"https:\u002F\u002Fjoin.slack.com\u002Ft\u002Fworldfoundrycommunity\u002Fshared_invite\u002Fzt-43nbi9fw4-okYiELzZHp0_1UPa3dh3bQ\">\n  \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSlack-Join%20Workspace-4A154B?logo=slack&logoColor=white\" alt=\"Join Slack\">\n\u003C\u002Fa>\n\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FWorldFoundry\u002FWorldFoundry\u002Fdiscussions\">\n  \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGitHub-Discussions-181717?logo=github\" alt=\"GitHub Discussions\">\n\u003C\u002Fa>\n\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Cstrong>WeChat Community\u003C\u002Fstrong>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Fwechat_qr.png\" alt=\"WorldFoundry WeChat Community\" width=\"280\">\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Cem>Scan the QR code to join our WeChat community.\u003Cbr>\n  The QR code is updated periodically if it expires.\u003C\u002Fem>\n\u003C\u002Fp>\n\n## 📰 News\n\n- **[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!\n- **[2026-07-11]** 🎉 **WorldFoundry is officially open-sourced.** We welcome ⭐ stars, bug reports, feature requests, and pull requests from the community!\n- **[Coming Soon]** Documentation improvements and additional benchmark integrations.\n\n\n![WorldFoundry teaser](docs\u002Ffumadocs\u002Fpublic\u002Fteaser.png)\n\n## Links\n\n- [Documentation](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Findex.mdx)\n- [Quickstart](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Fquickstart.mdx)\n- [Environment reference](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Freference\u002Fenvironments.mdx)\n- [Local asset preparation](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Fguides\u002Flocal-assets.mdx)\n- [TUI](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Fguides\u002Ftui.mdx)\n- [Inference guide](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Fguides\u002Finference.mdx)\n- [Studio guide](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Fguides\u002Fstudio.mdx)\n- [CLI reference](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Freference\u002Fcli.mdx)\n- [Supported models](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Fguides\u002Fsupported-models.mdx)\n- [Benchmark hub](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Fevaluation\u002Fbenchmark-hub\u002Findex.mdx)\n- [Contributing](CONTRIBUTING.md)\n\n## Demo Gallery\n\nThese examples are checked into the documentation site so a new user can see the expected artifact shape before running GPU jobs. Full release claims still require the matching run manifest, runtime profile, and validation scorecard.\n\n\u003Ctable>\n  \u003Ctr>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fltx2-3-i2v-penguin.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fltx2-3-i2v-penguin.gif\" width=\"100%\" alt=\"ltx2-3-i2v-penguin demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>LTX-2.3\u003C\u002Fstrong>\u003Cbr>\u003Csub>Image-to-video\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fwan2-1-vace-girl-snake.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fwan2-1-vace-girl-snake.gif\" width=\"100%\" alt=\"wan2-1-vace-girl-snake demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>Wan2.1 VACE\u003C\u002Fstrong>\u003Cbr>\u003Csub>Image\u002Fcontrol-to-video\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fskyreels-v3-reference-to-video.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fskyreels-v3-reference-to-video.gif\" width=\"100%\" alt=\"skyreels-v3-reference-to-video demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>SkyReels V3\u003C\u002Fstrong>\u003Cbr>\u003Csub>Reference-to-video\u003C\u002Fsub>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Funianimate-dit-human-animation.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Funianimate-dit-human-animation.gif\" width=\"100%\" alt=\"unianimate-dit-human-animation demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>UniAnimate-DiT\u003C\u002Fstrong>\u003Cbr>\u003Csub>Human animation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fopen-sora-plan-tokyo-street.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fopen-sora-plan-tokyo-street.gif\" width=\"100%\" alt=\"open-sora-plan-tokyo-street demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>Open-Sora-Plan\u003C\u002Fstrong>\u003Cbr>\u003Csub>Text-to-video\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fhunyuanvideo-i2v-firework-official.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fhunyuanvideo-i2v-firework-official.gif\" width=\"100%\" alt=\"hunyuanvideo-i2v-firework-official demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>HunyuanVideo I2V\u003C\u002Fstrong>\u003Cbr>\u003Csub>Image-to-video\u003C\u002Fsub>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fhunyuanvideo-t2v-cat-grass-official.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fhunyuanvideo-t2v-cat-grass-official.gif\" width=\"100%\" alt=\"hunyuanvideo-t2v-cat-grass-official demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>HunyuanVideo T2V\u003C\u002Fstrong>\u003Cbr>\u003Csub>Text-to-video\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fcogvideo_01.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fcogvideo_01.gif\" width=\"100%\" alt=\"cogvideo_01 demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>CogVideoX\u003C\u002Fstrong>\u003Cbr>\u003Csub>Text-to-video\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fac3d_02.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fac3d_02.gif\" width=\"100%\" alt=\"ac3d_02 demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>AC3D\u003C\u002Fstrong>\u003Cbr>\u003Csub>Camera\u002Fworld scene\u003C\u002Fsub>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fastra_02.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fastra_02.gif\" width=\"100%\" alt=\"astra_02 demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>Astra\u003C\u002Fstrong>\u003Cbr>\u003Csub>World navigation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fwarp_02.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fwarp_02.gif\" width=\"100%\" alt=\"warp_02 demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>Warp\u003C\u002Fstrong>\u003Cbr>\u003Csub>World navigation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fmatrix-game-2-official-universal.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fmatrix-game-2-official-universal.gif\" width=\"100%\" alt=\"matrix-game-2-official-universal demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>Matrix-Game-2\u003C\u002Fstrong>\u003Cbr>\u003Csub>Interactive world model\u003C\u002Fsub>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fhy-worldplay-official-8gpu.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fhy-worldplay-official-8gpu.gif\" width=\"100%\" alt=\"hy-worldplay-official-8gpu demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>HY-WorldPlay\u003C\u002Fstrong>\u003Cbr>\u003Csub>8-GPU image-pose world video\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fhunyuan-game-craft-village.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fhunyuan-game-craft-village.gif\" width=\"100%\" alt=\"hunyuan-game-craft-village demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>Hunyuan GameCraft\u003C\u002Fstrong>\u003Cbr>\u003Csub>Interactive village world\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fmatrix-game-3-cityscape.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fmatrix-game-3-cityscape.gif\" width=\"100%\" alt=\"matrix-game-3-cityscape demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>Matrix-Game-3\u003C\u002Fstrong>\u003Cbr>\u003Csub>Cityscape world model\u003C\u002Fsub>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fworldcam-industrial.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fworldcam-industrial.gif\" width=\"100%\" alt=\"worldcam-industrial demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>WorldCam\u003C\u002Fstrong>\u003Cbr>\u003Csub>Camera-path world video\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fyume-1p5-jungle-castle.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fyume-1p5-jungle-castle.gif\" width=\"100%\" alt=\"yume-1p5-jungle-castle demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>YUME-1.5\u003C\u002Fstrong>\u003Cbr>\u003Csub>First-person world navigation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fneoverse-robot-tabletop.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fneoverse-robot-tabletop.gif\" width=\"100%\" alt=\"neoverse-robot-tabletop demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>NeoVerse\u003C\u002Fstrong>\u003Cbr>\u003Csub>Robot video-input world model\u003C\u002Fsub>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fhunyuan-world-voyager-case1.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fhunyuan-world-voyager-case1.gif\" width=\"100%\" alt=\"hunyuan-world-voyager-case1 demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>HunyuanWorld-Voyager\u003C\u002Fstrong>\u003Cbr>\u003Csub>Conditioned world video\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fcosmos3.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fcosmos3.gif\" width=\"100%\" alt=\"cosmos3 demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>Cosmos3\u003C\u002Fstrong>\u003Cbr>\u003Csub>World video generation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fflashworld.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fflashworld.gif\" width=\"100%\" alt=\"flashworld demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>FlashWorld\u003C\u002Fstrong>\u003Cbr>\u003Csub>World video generation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fsana.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fsana.gif\" width=\"100%\" alt=\"sana demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>Sana\u003C\u002Fstrong>\u003Cbr>\u003Csub>Video generation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Flingbot-world.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Flingbot-world.gif\" width=\"100%\" alt=\"lingbot-world demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>LingBot World\u003C\u002Fstrong>\u003Cbr>\u003Csub>World-action generation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fwan2-2.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fwan2-2.gif\" width=\"100%\" alt=\"wan2-2 demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>Wan2.2\u003C\u002Fstrong>\u003Cbr>\u003Csub>Video generation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fluciddreamer.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fluciddreamer.gif\" width=\"100%\" alt=\"luciddreamer demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>LucidDreamer\u003C\u002Fstrong>\u003Cbr>\u003Csub>World video generation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fgen3c.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Fgen3c.gif\" width=\"100%\" alt=\"gen3c demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>GEN3C\u003C\u002Fstrong>\u003Cbr>\u003Csub>3D-aware video generation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n    \u003Ctd width=\"33%\">\n      \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry\u002Fraw\u002Fmain\u002Fdocs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Flongcat.mp4\">\u003Cimg src=\"docs\u002Ffumadocs\u002Fpublic\u002Freadme-demos\u002Flongcat.gif\" width=\"100%\" alt=\"longcat demo\">\u003C\u002Fa>\n      \u003Cbr>\u003Cstrong>LongCat\u003C\u002Fstrong>\u003Cbr>\u003Csub>World video generation\u003C\u002Fsub>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\nMore curated generated samples are embedded in the Studio docs.\n\n## What WorldFoundry Provides\n\n| Surface | Purpose | Entry point |\n| --- | --- | --- |\n| Model zoo | Catalogs video, world, 3D\u002F4D, VLA\u002FVA\u002FWAM, hosted API, and metadata-only model entries. | [`worldfoundry\u002Fdata\u002Fmodels\u002Fcatalog`](worldfoundry\u002Fdata\u002Fmodels\u002Fcatalog) |\n| In-tree runtimes | Keeps model architecture and inference adapters inside `worldfoundry`; checkpoints stay in local\u002FHugging Face caches. | [`worldfoundry\u002Fsynthesis`](worldfoundry\u002Fsynthesis), [`worldfoundry\u002Fpipelines`](worldfoundry\u002Fpipelines) |\n| TUI | Interactive model\u002Fbenchmark picker that prints runnable CLI commands. | `worldfoundry-eval tui` \u002F `worldfoundry-tui` |\n| Studio workspace | Browser UI for inference jobs, model-specific parameters, and artifact review. | [`worldfoundry.studio.workspace_app`](worldfoundry\u002Fstudio\u002Fworkspace_app.py) |\n| Benchmark zoo | Catalogs benchmark manifests, required assets, official runner constraints, and readiness states. | [`worldfoundry\u002Fdata\u002Fbenchmarks\u002Fcatalog`](worldfoundry\u002Fdata\u002Fbenchmarks\u002Fcatalog) |\n| Evaluation runner | Runs model × benchmark cells, imports existing outputs, and writes normalized scorecards. | [`worldfoundry\u002Fevaluation`](worldfoundry\u002Fevaluation) |\n| Docs | Bilingual Fumadocs site with setup, inference, evaluation, Studio, and maintainer guides. | [`docs\u002Ffumadocs`](docs\u002Ffumadocs) |\n\n## From Clone To First Run\n\nWorldFoundry uses conda as the supported open-source runtime path. Start with the unified GPU environment; only use a dedicated environment when a model profile documents a real ABI or simulator conflict. The full day-one path lives in the [Quickstart](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Fquickstart.mdx).\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FOpenEnvision\u002FWorldFoundry.git\ncd WorldFoundry\n\nbash scripts\u002Fsetup\u002Fbootstrap_worldfoundry.sh\nsource tmp\u002Fworldfoundry_unified_env.sh\nconda activate \"${WORLDFOUNDRY_UNIFIED_ENV_PREFIX}\"\n```\n\nCheckpoints, datasets, evaluator weights, API keys, and generated artifacts are **not** in git.\nSee [Local asset preparation](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Fguides\u002Flocal-assets.mdx) for cache layout, Hugging Face downloads, non-HF aliases, and benchmark assets.\n\nOn modern CUDA 12.8 hosts the installer resolves `worldfoundry-unified-cu128`. Pin a wheel tier only when the host requires it:\n\n```bash\nbash scripts\u002Fsetup\u002Fbootstrap_worldfoundry.sh --cuda cu124\nbash scripts\u002Fsetup\u002Fbootstrap_worldfoundry.sh --cuda cu121\n```\n\nKeep datasets and checkpoints outside the repository on shared machines:\n\n```bash\nbash scripts\u002Fsetup\u002Fbootstrap_worldfoundry.sh \\\n  --home \u002Fpath\u002Fto\u002Fworldfoundry-home \\\n  --data-root \u002Fpath\u002Fto\u002Fworldfoundry-data \\\n  --model-root \u002Fpath\u002Fto\u002Fworldfoundry-models \\\n  --artifact-root \u002Fpath\u002Fto\u002Fworldfoundry-artifacts\n```\n\nHugging Face models use native Hub loading (`from_pretrained`, `snapshot_download`, `HF_HOME` \u002F `HF_HUB_CACHE`, and `HF_TOKEN` for gated assets). `WORLDFOUNDRY_CKPT_DIR` remains for non-HF checkpoints and compatibility aliases.\n\nSome VLA\u002Faction policies need a documented model-specific environment (for example OpenVLA-OFT \u002F CogACT). Embodied simulator benchmarks follow the Docker VLA harness pattern — see the [environment reference](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Freference\u002Fenvironments.mdx).\n\nAfter the environment is active:\n\n```bash\nworldfoundry-eval --help\nworldfoundry-eval zoo models --json\nworldfoundry-eval zoo benchmarks --json\n```\n\n### Interactive first path (TUI)\n\n```bash\npython -m pip install -e \".[tui]\"\nworldfoundry-eval tui\n# or: worldfoundry-tui\n```\n\nThe TUI reads the same catalogs as the CLI and can print a runnable command before anything expensive runs:\n\n```bash\nworldfoundry-eval tui \\\n  --model-id \u003Cmodel-id> \\\n  --benchmark-id \u003Cbenchmark-id> \\\n  --print-command\n```\n\n### Scripted first model run\n\nPrepare assets, then launch a small demo. A common starter is `matrix-game-2` (public HF repo `Skywork\u002FMatrix-Game-2.0`):\n\n```bash\nbash scripts\u002Finference\u002Fprepare_model_infer.sh matrix-game-2 --download\nworldfoundry-eval zoo model-download --model-id matrix-game-2 --check-local --json\n\nbash scripts\u002Finference\u002Ftest_nav_video_gen.sh matrix-game-2 \\\n  --output-dir tmp\u002Fmatrix_game2_first_run\n```\n\nIf weights already live in a shared checkpoint tree, link them instead of copying:\n\n```bash\nbash scripts\u002Fsetup\u002Flink_hf_checkpoints.sh \\\n  --ckpt-dir \"${WORLDFOUNDRY_CKPT_DIR}\" \\\n  --hfd-root \"${WORLDFOUNDRY_HFD_ROOT}\" \\\n  --hf-hub-cache \"${HF_HUB_CACHE}\" \\\n  --default-world\n```\n\n## Run Inference\n\nPrefer the TUI or the documented inference helpers once assets are staged:\n\n```bash\nbash scripts\u002Finference\u002Ftest_nav_video_gen.sh matrix-game-2\n\nconda run -p \"${WORLDFOUNDRY_UNIFIED_ENV_PREFIX}\" \\\n  bash scripts\u002Finference\u002Ftest_nav_video_gen.sh matrix-game-2\n\nbash scripts\u002Finference\u002Frun_infer.sh --category video --model \u003Cmodel-id>\nbash scripts\u002Finference\u002Frun_infer.sh --category three_d_four_d --model \u003Cmodel-id>\n```\n\nCLI-shaped inference (same contract as Studio jobs):\n\n```bash\npython -m worldfoundry.studio.workspace_job infer \\\n  --model-id \u003Cmodel-id> \\\n  --prompt \"a cinematic scene, high quality\" \\\n  --output-dir tmp\u002Fworldfoundry_infer\u002F\u003Cmodel-id> \\\n  --device cuda\n```\n\nEach successful run should write media, logs, and manifest metadata under the output directory. Treat a file as demo evidence only after visual check and matching runtime-profile assumptions. Details: [Inference guide](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Fguides\u002Finference.mdx).\n\n## Launch Studio Workspace\n\nStudio is the preferred UI for release validation: model-specific forms, job status, preview media, and artifact links in one place. Start it from the same unified env used for inference:\n\n```bash\nsource tmp\u002Fworldfoundry_unified_env.sh\nconda activate \"${WORLDFOUNDRY_UNIFIED_ENV_PREFIX}\"\n\nbash scripts\u002Fworkspace\u002Frun_workspace.sh \\\n  --host 127.0.0.1 \\\n  --port 7870 \\\n  --max-jobs 8\n```\n\nOpen `http:\u002F\u002F127.0.0.1:7870\u002F`. If `python`, **LOAD**, or **START** fails with a missing interpreter, `cv2`, or `libssl`\u002F`libcrypto` error, recreate or verify the env and restart:\n\n```bash\nbash scripts\u002Fsetup\u002Fbootstrap_worldfoundry.sh --verify-only\nsource tmp\u002Fworldfoundry_unified_env.sh\nbash scripts\u002Fworkspace\u002Frun_workspace.sh\n```\n\nConfigure jobs in **Create Job**; optional shared defaults can use `WORLDFOUNDRY_STUDIO_SETTINGS_FILE`. Expensive runtime checks and preview builders are opt-in via `WORLDFOUNDRY_STUDIO_*` — see the [Studio guide](docs\u002Ffumadocs\u002Fcontent\u002Fdocs\u002Fguides\u002Fstudio.mdx).\n\nUse the **Visualizers** tab as the browser entrypoint for local preview services (World \u002F Gradio, Spark, Viser, Rerun, Embodied bridge). On a remote machine, forward port `7870` plus any viewer ports you launch.\n\nFor a single-model Studio process:\n\n```bash\nworldfoundry-studio\n```\n\n## Run Evaluation\n\nRun evaluation through a runnable benchmark path. Use `official-run` when the evaluator can execute locally; use `official-validation` when you already have official-shaped result files to import.\n\n```bash\nworldfoundry-eval run \\\n  --model matrix-game-2 \\\n  --benchmark vbench \\\n  --mode official-run \\\n  --output-dir tmp\u002Fhello_world_run \\\n  --json\n```\n\nInspect:\n\n- `run_manifest.json`: selected model, benchmark\u002Ftask metadata, timestamps, and output paths.\n- `results.jsonl`: per-sample generation records and artifact metadata.\n- `metrics\u002Fsummary.json`: aggregate metrics and failed\u002Fskipped sample counts.\n- `scorecard.json`: readiness, leaderboard eligibility, metric values, and blockers.\n\nFor model and benchmark discovery:\n\n```bash\nworldfoundry-eval tasks list\nworldfoundry-eval zoo models --json\nworldfoundry-eval zoo benchmarks --json\nworldfoundry-eval zoo model-show --model-id \u003Cmodel-id> --include-manifest --json\nworldfoundry-eval zoo benchmark-show --benchmark-id \u003Cbenchmark-id> --include-spec --json\n```\n\nFor existing official-shaped benchmark outputs:\n\n```bash\nworldfoundry-eval zoo benchmark-run \\\n  --benchmark-id vbench \\\n  --mode official-validation \\\n  --official-results-path \u003Cofficial_results.json> \\\n  --generated-artifact-dir \u003Cgenerated_videos> \\\n  --output-dir tmp\u002Fbenchmark_zoo\u002Fofficial_validation\u002Fvbench \\\n  --json\n```\n\nFor existing generated outputs:\n\n```bash\nworldfoundry-eval evaluate \\\n  --results-path tmp\u002Fresults.jsonl \\\n  --output-dir tmp\u002Fworldfoundry_evaluate \\\n  --metric artifact_count \\\n  --required-artifact video \\\n  --json\n```\n\nFor the formal benchmark inventory, review the expanded plan first:\n\n```bash\nworldfoundry-eval prepare \\\n  --all-benchmarks \\\n  --output-dir tmp\u002Fworldfoundry_all_benchmarks_plan \\\n  --json\n\nworldfoundry-eval run \\\n  --all-benchmarks \\\n  --model \u003Cmodel-zoo-id> \\\n  --prepare \\\n  --data-root cache\u002Fworldfoundry\u002Fdata\u002Fhfd_datasets \\\n  --plan-only \\\n  --output-dir tmp\u002Fworldfoundry_all_benchmarks_plan \\\n  --json\n```\n\nUse the integrity commands before claiming benchmark support:\n\n```bash\nworldfoundry-eval zoo benchmarks --json\nworldfoundry-eval run --plan-only --json\n```\n\nFor release audits, use public CLI surfaces only:\n\n```bash\nworldfoundry-eval validate-artifact --json\n```\n\nContract runs, normalizer-only imports, partial dataset runs, and missing-official-runner checks are not leaderboard evidence. A public claim needs the full official data\u002Fruntime path and a scorecard whose eligibility fields explicitly support the claim.\n\n## Documentation Site\n\nRun the docs locally:\n\n```bash\ncd docs\u002Ffumadocs\nnpm ci\nnpm run dev -- --port 8014\n```\n\nBuild the static docs from the repository root:\n\n```bash\nbash scripts\u002Fdocs\u002Fbuild.sh\n```\n\nThe docs app serves English routes under `\u002Fdocs` and Chinese routes under `\u002Fzh\u002Fdocs`.\n\n## Development Checks\n\nUse these checks before opening a PR or marking a model\u002Fbenchmark ready:\n\n```bash\nsource tmp\u002Fworldfoundry_unified_env.sh\nconda activate \"${WORLDFOUNDRY_UNIFIED_ENV_PREFIX}\"\n\nPYTHONPATH=. python -m compileall -q worldfoundry scripts\nPYTHONPATH=. python -m pytest -m fast_eval_core test\u002Feval_core\nbash scripts\u002Fdocs\u002Fbuild.sh --skip-bootstrap\n\nworldfoundry-eval zoo model-download --model-id \u003Cmodel-id> --check-local --json\nworldfoundry-eval zoo benchmark-download --benchmark-id \u003Cbenchmark-id> --check-local --json\nworldfoundry-eval run --plan-only --fail-on-overclaim --fail-on-stale --json\n```\n\nWhen adding or changing a model:\n\n1. Port required inference code into `worldfoundry`; do not depend on a cloned external repo at runtime.\n2. Keep official repositories only as provenance or parity references.\n3. Declare checkpoints, runtime variables, and environment assumptions in the model catalog\u002Fruntime profile.\n4. Run the smallest official-style demo and visually inspect the artifact.\n5. Record evidence in the docs before promoting readiness.\n\n## Repository Layout\n\n```text\nWorldFoundry\n├─ docs\u002Ffumadocs                         # Documentation site, teaser, screenshots, and demo media\n├─ requirements                          # Unified and optional dependency presets\n├─ scripts\n│  ├─ inference                          # User-facing inference entrypoints\n│  ├─ setup                              # Conda setup wrappers\n│  ├─ workspace                          # Studio \u002F Workspace launch helpers\n│  └─ docs                               # Documentation build wrapper\n├─ worldfoundry\n│  ├─ core                               # Shared contracts and reusable runtime abstractions\n│  ├─ data                               # Model\u002Fbenchmark catalogs, runtime profiles, fixtures\n│  ├─ evaluation                         # Runner, tasks, metrics, scorecards, reports\n│  ├─ operators                          # Input validation, preprocessing, interaction handling\n│  ├─ pipelines                          # User-facing pipeline wrappers\n│  ├─ representations                    # 3D\u002F4D and spatial representation outputs\n│  ├─ runtime                            # Runtime paths, assets, jobs, and probes\n│  ├─ studio                             # Workspace and Studio frontends\n│  └─ synthesis                          # In-tree model synthesis\u002Faction-generation runtimes\n├─ test                                  # Test suites\n├─ thirdparty                            # Reviewed vendored\u002Fnative dependencies\n└─ tools                                 # Maintenance and asset utilities\n```\n\n## Citation\n\nIf you use WorldFoundry or its benchmark\u002Fmodel integrations in research, cite this repository and the upstream methods, checkpoints, datasets, and benchmarks that your run depends on. A formal paper citation will be added when the technical report is released.\n\n## Acknowledgment\n\nWorldFoundry integrates and wraps a large set of upstream world-model, video-generation, perception, reconstruction, and embodied-action projects. See the method-specific runtime profiles and the docs appendix for upstream project pointers and licenses.\n\nWe also thank the following open-source projects for their design, runtime, and evaluation ideas that informed WorldFoundry:\n\n- [FlashDreams](https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fflashdreams) — high-performance inference and serving for interactive autoregressive video and world models\n- [FastVideo](https:\u002F\u002Fgithub.com\u002Fhao-ai-lab\u002FFastVideo) — a unified inference and post-training framework for accelerated video generation\n- [OpenWorldLib](https:\u002F\u002Fgithub.com\u002FOpenDCAI\u002FOpenWorldLib) — a unified codebase for advanced world models\n- [VLA Evaluation Harness](https:\u002F\u002Fgithub.com\u002Fallenai\u002Fvla-evaluation-harness) — one framework to evaluate VLA models on robot simulation benchmarks\n","WorldFoundry 是一个面向世界模型（World Model）的开源推理与评估基础设施，提供统一的运行时栈，支持视频生成、3D\u002F4D表征、具身动作及交互式世界等多模态任务的本地化推理（含TUI\u002FCLI\u002FStudio界面）和标准化基准评测。其核心特点包括资产与环境分离管理、模块化推理流水线、可复现的评估协议（基于预定义benchmark布局与scorecard机制），以及面向研究与工程验证的一体化工作流。适用于AI基础模型研发团队、具身智能实验室及多模态世界建模方向的研究者开展模型迭代、跨任务横向对比与部署前合规性验证。",2,"2026-07-12 02:30:02","CREATED_QUERY"]