{"id":50535224,"url":"https://github.com/nvidia/flashdreams","last_synced_at":"2026-06-03T16:00:48.463Z","repository":{"id":362214286,"uuid":"1218835372","full_name":"NVIDIA/flashdreams","owner":"NVIDIA","description":"high-performance inference and serving library for interactive autoregressive video and world models","archived":false,"fork":false,"pushed_at":"2026-06-03T07:04:43.000Z","size":161261,"stargazers_count":26,"open_issues_count":30,"forks_count":2,"subscribers_count":1,"default_branch":"main","last_synced_at":"2026-06-03T07:27:30.124Z","etag":null,"topics":["efficiency","interactive","video-models","world-models"],"latest_commit_sha":null,"homepage":"https://nvidia.github.io/flashdreams/","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"other","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/NVIDIA.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":"CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":"SECURITY.md","support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":"NOTICE","maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-04-23T09:03:25.000Z","updated_at":"2026-06-03T07:03:47.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/NVIDIA/flashdreams","commit_stats":null,"previous_names":["nvidia/flashdreams"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/NVIDIA/flashdreams","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/NVIDIA%2Fflashdreams","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/NVIDIA%2Fflashdreams/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/NVIDIA%2Fflashdreams/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/NVIDIA%2Fflashdreams/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/NVIDIA","download_url":"https://codeload.github.com/NVIDIA/flashdreams/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/NVIDIA%2Fflashdreams/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":33872298,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-03T02:00:06.370Z","response_time":59,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["efficiency","interactive","video-models","world-models"],"created_at":"2026-06-03T16:00:37.895Z","updated_at":"2026-06-03T16:00:48.457Z","avatar_url":"https://github.com/NVIDIA.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003c!--\nSPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION \u0026 AFFILIATES. All rights reserved.\nSPDX-License-Identifier: Apache-2.0\n--\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003cpicture\u003e\n    \u003csource media=\"(prefers-color-scheme: dark)\" srcset=\"assets/logo/horizontal-dark.svg\"\u003e\n    \u003cimg alt=\"FlashDreams\" src=\"assets/logo/horizontal-light.svg\" width=\"600\"\u003e\n  \u003c/picture\u003e\n\u003c/p\u003e\n\n\u003cp align=\"center\"\u003e\n  \u003ca href=\"LICENSE\"\u003e\u003cimg alt=\"License: Apache 2.0\" src=\"https://img.shields.io/badge/License-Apache_2.0-blue.svg\"\u003e\u003c/a\u003e\n  \u003ca href=\"https://nvidia.github.io/flashdreams/main/index.html\"\u003e\u003cimg alt=\"Documentation\" src=\"https://img.shields.io/badge/docs-latest-blue.svg\"\u003e\u003c/a\u003e\n\u003c/p\u003e\n\n**FlashDreams** is a high-performance inference and serving library for\ninteractive autoregressive video and world models. It began as the optimized\nruntime behind the [NVIDIA OmniDreams closed-loop demo for GTC 2026][omnidreams-blog]\nand has grown into a general platform for real-time world-model applications\nacross gaming, autonomous vehicles, robotics, simulated or virtual\nenvironments, and more.\n\n[omnidreams-blog]: https://research.nvidia.com/labs/sil/projects/omnidreams-blog/\n\nhttps://github.com/user-attachments/assets/2b000ce9-effe-4cc9-a227-5b4619413e4d\n\n## System Requirements\n\n- NVIDIA GPU with **80 GB VRAM or more** (e.g. H100 80GB), see notes below.\n- NVIDIA driver from the **R580 series or newer** (compatible with CUDA 13.x)\n- **CUDA 13.x** (PyTorch `2.11.0+cu130` and the `nvidia-*-cu13` libraries are\n  resolved by `uv sync`. A system CUDA toolkit is needed only for the\n  developer extras and is included in `nvidia/cuda:13.2.1-cudnn-devel-ubuntu24.04`)\n- **Python \u003e= 3.10**\n- **PyTorch \u003e= 2.11.0+cu130** (`\u003e= 2.9` for bare PyPI library install)\n- Linux x86-64 or arm64\n- **100 GB+ free storage space** recommended for environment and model checkpoints.\n- Docker with the\n  [NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html)\n  (optional, only for the container workflow)\n\n\u003e Development and testing were performed on GPUs with **80 GB of VRAM or more**.\n\u003e Inference can fail (out-of-memory) on consumer and even enthusiast GPUs.\n\u003e Per-model GPU and VRAM requirements are listed on each model page in\n\u003e [the model gallery](https://nvidia.github.io/flashdreams/main/models/index.html).\n\n## Quickstart\n\nThe complete setup is in\n[the installation guide](https://nvidia.github.io/flashdreams/main/quickstart/installation.html).\nAssuming `uv` is [installed](https://docs.astral.sh/uv/getting-started/installation), the shortest viable path is:\n\n```bash\ngit clone https://github.com/NVIDIA/flashdreams.git\ncd flashdreams\nuv sync --extra runners\nexport HF_TOKEN=\u003cyour-hf-token\u003e\nuv run flashdreams-run --help\n```\n\nNote for developers/maintainers you would want to run `uv sync --extra dev --extra runners` instead.\n\nThen launch your first model by following\n[the quickstart guide](https://nvidia.github.io/flashdreams/main/quickstart/first_world_model.html).\nFor example, the offline Self-Forcing T2V quickstart command is:\n\n```bash\nuv run --project integrations/self_forcing \\\n    flashdreams-run self-forcing-wan2.1-t2v-1.3b \\\n    --total-blocks 7\n```\n\nYou can also install FlashDreams as a library from PyPI:\n\n```bash\npip install flashdreams\n```\n\n## Supported models\n\nFlashDreams ships first-party integrations under\n[`integrations/`](integrations/). Each model has a dedicated docs page with\nrunner slugs, multi-GPU commands, and (where available) profiling benchmarks.\n\n| Model | Family |\n| --- | --- |\n| [Self-Forcing](https://nvidia.github.io/flashdreams/main/models/self_forcing.html) | Streaming Wan2.1 T2V |\n| [OmniDreams](https://nvidia.github.io/flashdreams/main/models/omnidreams.html) | HDMap-conditioned driving world model |\n| [LingBot-World](https://nvidia.github.io/flashdreams/main/models/lingbot_world.html) | Camera-controllable I2V world model |\n| [Wan2.1](https://nvidia.github.io/flashdreams/main/models/wan21.html) | Bidirectional T2V / I2V |\n| [Causal-Forcing](https://nvidia.github.io/flashdreams/main/models/causal_forcing.html) | Streaming Wan2.1 T2V / I2V |\n| [Causal Wan2.2](https://nvidia.github.io/flashdreams/main/models/causal_wan22.html) | FastVideo Causal Wan 2.2 14B MoE T2V |\n| [FlashVSR](https://nvidia.github.io/flashdreams/main/models/flashvsr.html) | Streaming video super-resolution |\n| [Cosmos-Predict2.5](https://nvidia.github.io/flashdreams/main/models/cosmos_predict2.html) | Bidirectional T2V / I2V |\n\nSee [the model gallery](https://nvidia.github.io/flashdreams/main/models/index.html) and\n[the new method guide](https://nvidia.github.io/flashdreams/main/developer_guides/new_integration.html)\nto add your own.\n\n## Developer guides\n\n- [Inference pipeline overview](https://nvidia.github.io/flashdreams/main/developer_guides/inference_pipeline_overview.html)\n- [Config system](https://nvidia.github.io/flashdreams/main/developer_guides/config_system.html)\n- [Add a new method](https://nvidia.github.io/flashdreams/main/developer_guides/new_integration.html)\n\nFor day-to-day development:\n\n```bash\nuv sync --extra dev --extra runners\nuv run pre-commit run -a\nuv run pytest -m \"not manual\"\n```\n\nSee [`DEV.md`](DEV.md) for repository-specific workflow notes.\n\n## Contributing\n\nFor how to contribute, see [`CONTRIBUTING.md`](CONTRIBUTING.md).\nNew integrations, bug reports, feature requests, performance tuning, and\ndocumentation edits are all welcome.\n\nUse [GitHub Issues](https://github.com/NVIDIA/flashdreams/issues) to report defects or request improvements.\n\nJoin us on the [NVIDIA Omniverse Discord](https://discord.com/invite/nvidiaomniverse)\nto share your results and take part in technical discussion! Channel: [`#flashdreams`](https://discord.gg/yTdHDqFP)\n\n## Security\n\nTo report a potential security vulnerability, follow the coordinated\ndisclosure process in [`SECURITY.md`](SECURITY.md).\n\n## License\n\nFlashDreams is released under the [Apache License 2.0](LICENSE). Third-party\ncomponents and their licenses are listed in\n[`THIRD-PARTY-NOTICES`](THIRD-PARTY-NOTICES) and [`NOTICE`](NOTICE). The\nrepository is REUSE-compliant; see [`REUSE.toml`](REUSE.toml) and\n[`LICENSES/`](LICENSES/).\n\n## Citation\n\nIf FlashDreams is useful in your research or product, please cite the project:\n\n```bibtex\n@misc{flashdreams2026,\n  title        = {FlashDreams: High-performance inference and serving for\n                  interactive autoregressive video and world models},\n  author       = {{FlashDreams Contributors}},\n  year         = {2026},\n  howpublished = {\\url{https://github.com/NVIDIA/flashdreams}},\n}\n\n@misc{nvidia2026omnidreams,\n  title={OmniDreams: Real-Time Generative Closed-Loop Autonomous Vehicle Simulation Built on NVIDIA Cosmos},\n  author={Basant, Aarti and Kar, Amlan and Paschalidou, Despoina and Garcia Cobo, Guillermo and Turki, Haithem and Ling, Huan and Seo, Jaewoo and Wang, Jialiang and Lucas, James and Wu, Jay and Lorraine, Jonathan and Gao, Jun and He, Kai and Tothova, Katarina and Xie, Kevin and Tyszkiewicz, Michal and Wu, Qi and de Lutio, Riccardo and Li, Ruilong and Fidler, Sanja and Kim, Seung Wook and Shen, Tianchang and Cao, Tianshi and Pfaff, Tobias and Lew, William and Ren, Xuanchi and Lu, Yifan and Gojcic, Zan and Wang, Zian},\n  year={2026},\n  note={Technical report},\n  howpublished={\\url{https://research.nvidia.com/labs/sil/projects/omnidreams-blog/paper.pdf}}\n}\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fnvidia%2Fflashdreams","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fnvidia%2Fflashdreams","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fnvidia%2Fflashdreams/lists"}