[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-Robbyant--lingbot-world":3,"tool-Robbyant--lingbot-world":64},[4,17,26,40,48,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,2,"2026-04-03T11:11:01",[13,14,15],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":23,"last_commit_at":32,"category_tags":33,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,34,35,36,15,37,38,13,39],"数据工具","视频","插件","其他","语言模型","音频",{"id":41,"name":42,"github_repo":43,"description_zh":44,"stars":45,"difficulty_score":10,"last_commit_at":46,"category_tags":47,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,38,37],{"id":49,"name":50,"github_repo":51,"description_zh":52,"stars":53,"difficulty_score":10,"last_commit_at":54,"category_tags":55,"status":16},519,"PaddleOCR","PaddlePaddle\u002FPaddleOCR","PaddleOCR 是一款基于百度飞桨框架开发的高性能开源光学字符识别工具包。它的核心能力是将图片、PDF 等文档中的文字提取出来，转换成计算机可读取的结构化数据，让机器真正“看懂”图文内容。\n\n面对海量纸质或电子文档，PaddleOCR 解决了人工录入效率低、数字化成本高的问题。尤其在人工智能领域，它扮演着连接图像与大型语言模型（LLM）的桥梁角色，能将视觉信息直接转化为文本输入，助力智能问答、文档分析等应用场景落地。\n\nPaddleOCR 适合开发者、算法研究人员以及有文档自动化需求的普通用户。其技术优势十分明显：不仅支持全球 100 多种语言的识别，还能在 Windows、Linux、macOS 等多个系统上运行，并灵活适配 CPU、GPU、NPU 等各类硬件。作为一个轻量级且社区活跃的开源项目，PaddleOCR 既能满足快速集成的需求，也能支撑前沿的视觉语言研究，是处理文字识别任务的理想选择。",74913,"2026-04-05T10:44:17",[38,14,13,37],{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":23,"last_commit_at":62,"category_tags":63,"status":16},2471,"tesseract","tesseract-ocr\u002Ftesseract","Tesseract 是一款历史悠久且备受推崇的开源光学字符识别（OCR）引擎，最初由惠普实验室开发，后由 Google 维护，目前由全球社区共同贡献。它的核心功能是将图片中的文字转化为可编辑、可搜索的文本数据，有效解决了从扫描件、照片或 PDF 文档中提取文字信息的难题，是数字化归档和信息自动化的重要基础工具。\n\n在技术层面，Tesseract 展现了强大的适应能力。从版本 4 开始，它引入了基于长短期记忆网络（LSTM）的神经网络 OCR 引擎，显著提升了行识别的准确率；同时，为了兼顾旧有需求，它依然支持传统的字符模式识别引擎。Tesseract 原生支持 UTF-8 编码，开箱即用即可识别超过 100 种语言，并兼容 PNG、JPEG、TIFF 等多种常见图像格式。输出方面，它灵活支持纯文本、hOCR、PDF、TSV 等多种格式，方便后续数据处理。\n\nTesseract 主要面向开发者、研究人员以及需要构建文档处理流程的企业用户。由于它本身是一个命令行工具和库（libtesseract），不包含图形用户界面（GUI），因此最适合具备一定编程能力的技术人员集成到自动化脚本或应用程序中",73286,"2026-04-03T01:56:45",[13,14],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":75,"owner_avatar_url":76,"owner_bio":77,"owner_company":78,"owner_location":78,"owner_email":78,"owner_twitter":79,"owner_website":80,"owner_url":81,"languages":82,"stars":87,"forks":88,"last_commit_at":89,"license":90,"difficulty_score":10,"env_os":91,"env_gpu":92,"env_ram":91,"env_deps":93,"category_tags":100,"github_topics":101,"view_count":106,"oss_zip_url":78,"oss_zip_packed_at":78,"status":16,"created_at":107,"updated_at":108,"faqs":109,"releases":150},660,"Robbyant\u002Flingbot-world","lingbot-world","Advancing Open-source World Models","LingBot-World 是由 Robbyant Team 推出的开源世界模拟器，旨在通过视频生成技术构建高保真的虚拟环境。它主要解决了现有视频模型在长时序一致性差、交互响应慢方面的痛点。\n\nLingBot-World 不仅能在写实、科学场景及卡通风格等多种环境中保持逼真的动态效果，还具备独特的“长时记忆”能力，能维持分钟级别的上下文一致性。更难得的是，它实现了实时交互，生成 16 帧画面的延迟控制在 1 秒以内，大大提升了实用性。\n\nLingBot-World 非常适合开发者、AI 研究人员以及内容创作者使用。无论是用于游戏原型设计、机器人训练，还是影视特效制作，用户都可以直接获取其开源代码和模型权重。作为连接开源与闭源技术的桥梁，它为社区提供了强大的底层能力，助力更多创新应用落地。目前基础版本已支持相机姿态控制与动作指令输入，欢迎体验。","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FRobbyant_lingbot-world_readme_71965c487c3d.png\">\n\n\u003Ch1>LingBot-World: Advancing Open-source World Models\u003C\u002Fh1>\n\nRobbyant Team\n\n\u003C\u002Fdiv>\n\n\n\u003Cdiv align=\"center\">\n\n[![Page](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%8C%90%20Project%20Page-Demo-00bfff)](https:\u002F\u002Ftechnology.robbyant.com\u002Flingbot-world)\n[![Tech Report](https:\u002F\u002Fimg.shields.io\u002Fstatic\u002Fv1?label=Paper&message=PDF&color=red&logo=arxiv)](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.20540)\n[![Model](https:\u002F\u002Fimg.shields.io\u002Fstatic\u002Fv1?label=%F0%9F%A4%97%20Model&message=HuggingFace&color=yellow)](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Frobbyant\u002Flingbot-world)\n[![Model](https:\u002F\u002Fimg.shields.io\u002Fstatic\u002Fv1?label=%F0%9F%A4%96%20Model&message=ModelScope&color=purple)](https:\u002F\u002Fwww.modelscope.cn\u002Fmodels\u002FRobbyant\u002Flingbot-world-base-cam)\n[![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache--2.0-green)](LICENSE.txt)\n\n\n\u003C\u002Fdiv>\n\n-----\n\nWe are excited to introduce **LingBot-World**, an open-sourced world simulator stemming from video generation. Positioned\nas a top-tier world model, LingBot-World offers the following features. \n- **High-Fidelity & Diverse Environments**: It maintains high fidelity and robust dynamics in a broad spectrum of environments, including realism, scientific contexts, cartoon styles, and beyond. \n- **Long-Term Memory & Consistency**: It enables a minute-level horizon while preserving contextual consistency over time, which is also known as long-term memory. \n- **Real-Time Interactivity & Open Access**: It supports real-time interactivity, achieving a latency of under 1 second when producing 16 frames per second. We provide public access to the code and model in an effort to narrow the divide between open-source and closed-source technologies. We believe our release will empower the community with practical applications across areas like content creation, gaming, and robot learning.\n\n## 🎬 Video Demo\n\u003Cdiv align=\"center\">\n  \u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fea4a7a8d-5d9e-4ccf-96e7-02f93797116e\" width=\"100%\" poster=\"\"> \u003C\u002Fvideo>\n\u003C\u002Fdiv>\n\n## 🔥 News\n- Mar 2, 2026: 🎉 We release the **LingBot-World-Base (Act)** model weights.\n- Jan 29, 2026: 🎉 We release the technical report, code, and models for LingBot-World.\n\n\u003C!-- ## 🔖 Introduction of LingBot-World\nWe present **LingBot-World**, an **open-sourced** world simulator stemming from video generation. Positioned\nas a top-tier world model, LingBot-World offers the following features. \n- It maintains high fidelity and robust dynamics in a broad spectrum of environments, including realism, scientific contexts, cartoon styles, and beyond. \n- It enables a minute-level horizon while preserving contextual consistency over time, which is also known as **long-term memory**. \n- It supports real-time interactivity, achieving a latency of under 1 second when producing 16 frames per second. We provide public access to the code and model in an effort to narrow the divide between open-source and closed-source technologies. We believe our release will empower the community with practical applications across areas like content creation, gaming, and robot learning. -->\n\n## ⚙️ Quick Start\nThis codebase is built upon [Wan2.2](https:\u002F\u002Fgithub.com\u002FWan-Video\u002FWan2.2). Please refer to their documentation for installation instructions.\n### Installation\nClone the repo:\n```sh\ngit clone https:\u002F\u002Fgithub.com\u002Frobbyant\u002Flingbot-world.git\ncd lingbot-world\n```\nInstall dependencies:\n```sh\n# Ensure torch >= 2.4.0\npip install -r requirements.txt\n```\nInstall [`flash_attn`](https:\u002F\u002Fgithub.com\u002FDao-AILab\u002Fflash-attention):\n```sh\npip install flash-attn --no-build-isolation\n```\n### Model Download\n\n| Model | Control Signals | Resolution | Download Links |\n| :---  | :--- | :--- | :--- |\n| **LingBot-World-Base (Cam)** | Camera Poses | 480P & 720P | 🤗 [HuggingFace](https:\u002F\u002Fhuggingface.co\u002Frobbyant\u002Flingbot-world-base-cam) 🤖 [ModelScope](https:\u002F\u002Fwww.modelscope.cn\u002Fmodels\u002FRobbyant\u002Flingbot-world-base-cam) |\n| **LingBot-World-Base (Act)** | Actions | 480P & 720P | 🤗 [HuggingFace](https:\u002F\u002Fhuggingface.co\u002Frobbyant\u002Flingbot-world-base-act) |\n| **LingBot-World-Fast**       |    -    | - | *To be released* |\n\nDownload models using huggingface-cli:\n```sh\npip install \"huggingface_hub[cli]\"\nhuggingface-cli download robbyant\u002Flingbot-world-base-cam --local-dir .\u002Flingbot-world-base-cam\n```\nDownload models using modelscope-cli:\n ```sh\npip install modelscope\nmodelscope download robbyant\u002Flingbot-world-base-cam --local_dir .\u002Flingbot-world-base-cam\n```\n### Inference\nBefore running inference, you need to prepare:\n- Input image\n- Text prompt\n- Control signals (optional, can be generated from a video using [ViPE](https:\u002F\u002Fgithub.com\u002Fnv-tlabs\u002Fvipe))\n  - `intrinsics.npy`: Shape `[num_frames, 4]`, where the 4 values represent `[fx, fy, cx, cy]`\n  - `poses.npy`: Shape `[num_frames, 4, 4]`, where each `[4, 4]` represents a transformation matrix in OpenCV coordinates\n\nWe provide the following reference inference scripts:\n- `LingBot-World-Base (Cam)`:\n  - 480P:\n  ``` sh\n  torchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 480*832 --ckpt_dir lingbot-world-base-cam --image examples\u002F00\u002Fimage.jpg --action_path examples\u002F00 --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt \"The video presents a soaring journey through a fantasy jungle. The wind whips past the rider's blue hands gripping the reins, causing the leather straps to vibrate. The ancient gothic castle approaches steadily, its stone details becoming clearer against the backdrop of floating islands and distant waterfalls.\"\n  ```\n  - 720P:\n  ``` sh\n  torchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 720*1280 --ckpt_dir lingbot-world-base-cam --image examples\u002F00\u002Fimage.jpg --action_path examples\u002F00 --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt \"The video presents a soaring journey through a fantasy jungle. The wind whips past the rider's blue hands gripping the reins, causing the leather straps to vibrate. The ancient gothic castle approaches steadily, its stone details becoming clearer against the backdrop of floating islands and distant waterfalls.\"\n  ```\n  Alternatively, you can run inference without control signals:\n  ``` sh\n  torchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 480*832 --ckpt_dir lingbot-world-base-cam --image examples\u002F00\u002Fimage.jpg --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt \"The video presents a soaring journey through a fantasy jungle. The wind whips past the rider's blue hands gripping the reins, causing the leather straps to vibrate. The ancient gothic castle approaches steadily, its stone details becoming clearer against the backdrop of floating islands and distant waterfalls.\"\n  ```\n- `LingBot-World-Base (Act)`:\n  - 480P:\n  ``` sh\n  torchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 480*832 --ckpt_dir lingbot-world-base-act --image examples\u002F00\u002Fimage.jpg --action_path examples\u002F00 --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt \"The video presents a soaring journey through a fantasy jungle. The wind whips past the rider's blue hands gripping the reins, causing the leather straps to vibrate. The ancient gothic castle approaches steadily, its stone details becoming clearer against the backdrop of floating islands and distant waterfalls.\"\n  ```\n  - 720P:\n  ``` sh\n  torchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 720*1280 --ckpt_dir lingbot-world-base-act --image examples\u002F00\u002Fimage.jpg --action_path examples\u002F00 --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt \"The video presents a soaring journey through a fantasy jungle. The wind whips past the rider's blue hands gripping the reins, causing the leather straps to vibrate. The ancient gothic castle approaches steadily, its stone details becoming clearer against the backdrop of floating islands and distant waterfalls.\"\n  ```\nTips:\nIf you have sufficient CUDA memory, you may increase the `frame_num` parameter to a value such as 961 to generate a one-minute video at 16 FPS. Otherwise if the CUDA memory is not sufficient, you may use ``--t5_cpu`` to decrease the memory usage.\n\n### Quantized Model for Limited GPU Resources\nWe sincerely thank the community for their valuable support and contributions in LingBot-World. For users with limited GPU memory, we recommend using a **4-bit quantized version** of LingBot-World-Base (Cam), which significantly reduces GPU memory consumption while maintaining competitive visual quality for inference.\n\n👉 Download link: https:\u002F\u002Fhuggingface.co\u002Fcahlen\u002Flingbot-world-base-cam-nf4\n\n> ⚠️ Note: This quantized model is intended **for inference only**. Minor degradation in visual fidelity and temporal consistency may occur compared to the full-precision model.\n\n### Demo Results\nWe provide comparison demos where camera parameters are estimated by [ViPE](https:\u002F\u002Fgithub.com\u002Fnv-tlabs\u002Fvipe) from original videos downloaded from [Genie3](https:\u002F\u002Fdeepmind.google\u002Fblog\u002Fgenie-3-a-new-frontier-for-world-models\u002F):\n\u003Cdiv align=\"center\">\n  \u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Ffc95ee9e-e8a9-4f70-9aa2-9536c8365ccc\" width=\"100%\" poster=\"\"> \u003C\u002Fvideo>\n  \u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fbac89021-b394-4f68-a688-9a0b90e30241\" width=\"100%\" poster=\"\"> \u003C\u002Fvideo>\n\u003C\u002Fdiv>\n\n## 📚 Related Projects\n- [HoloCine](https:\u002F\u002Fholo-cine.github.io\u002F)\n- [Ditto](https:\u002F\u002Fezioby.github.io\u002FDitto_page\u002F)\n- [WorldCanvas](https:\u002F\u002Fworldcanvas.github.io\u002F)\n- [RewardForcing](https:\u002F\u002Freward-forcing.github.io\u002F)\n- [CoDeF](https:\u002F\u002Fqiuyu96.github.io\u002FCoDeF\u002F)\n\n## 📜 License\nThis project is licensed under the Apache 2.0 License. Please refer to the [LICENSE file](LICENSE.txt) for the full text, including details on rights and restrictions.\n\n## ✨ Acknowledgement\nWe would like to express our gratitude to the Wan Team for open-sourcing their code and models. Their contributions have been instrumental to the development of this project.\n\n## 📖 Citation\nIf you find this work useful for your research, please cite our paper:\n\n```\n@article{lingbot-world,\n      title={Advancing Open-source World Models}, \n      author={Robbyant Team and Zelin Gao and Qiuyu Wang and Yanhong Zeng and Jiapeng Zhu and Ka Leong Cheng and Yixuan Li and Hanlin Wang and Yinghao Xu and Shuailei Ma and Yihang Chen and Jie Liu and Yansong Cheng and Yao Yao and Jiayi Zhu and Yihao Meng and Kecheng Zheng and Qingyan Bai and Jingye Chen and Zehong Shen and Yue Yu and Xing Zhu and Yujun Shen and Hao Ouyang},\n      journal={arXiv preprint arXiv:2601.20540},\n      year={2026}\n}\n```\n","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FRobbyant_lingbot-world_readme_71965c487c3d.png\">\n\n\u003Ch1>LingBot-World：推动开源世界模型发展\u003C\u002Fh1>\n\nRobbyant 团队\n\n\u003C\u002Fdiv>\n\n\n\u003Cdiv align=\"center\">\n\n[![Page](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%8C%90%20Project%20Page-Demo-00bfff)](https:\u002F\u002Ftechnology.robbyant.com\u002Flingbot-world)\n[![Tech Report](https:\u002F\u002Fimg.shields.io\u002Fstatic\u002Fv1?label=Paper&message=PDF&color=red&logo=arxiv)](https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.20540)\n[![Model](https:\u002F\u002Fimg.shields.io\u002Fstatic\u002Fv1?label=%F0%9F%A4%97%20Model&message=HuggingFace&color=yellow)](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Frobbyant\u002Flingbot-world)\n[![Model](https:\u002F\u002Fimg.shields.io\u002Fstatic\u002Fv1?label=%F0%9F%A4%96%20Model&message=ModelScope&color=purple)](https:\u002F\u002Fwww.modelscope.cn\u002Fmodels\u002FRobbyant\u002Flingbot-world-base-cam)\n[![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache--2.0-green)](LICENSE.txt)\n\n\n\u003C\u002Fdiv>\n\n-----\n\n我们很高兴推出 **LingBot-World**，这是一个源自视频生成的开源世界模拟器 (World Simulator)。作为顶级世界模型 (World Model)，LingBot-World 提供以下功能。 \n- **高保真与多样化环境**：它在广泛的环境（包括写实、科学场景、卡通风格等）中保持高保真度和稳健的动态效果。 \n- **长期记忆与一致性**：它实现了分钟级的时间跨度，同时保持上下文随时间的一致性，这也被称为长期记忆 (Long-Term Memory)。 \n- **实时交互与开放访问**：它支持实时交互，在生成每秒 16 帧 (Frames Per Second) 时延迟低于 1 秒。我们公开提供了代码和模型的访问权限，旨在缩小开源 (Open-Source) 与闭源技术之间的差距。我们相信我们的发布将为社区赋能，使其在内容创作、游戏和机器人学习 (Robot Learning) 等领域获得实际应用。\n\n## 🎬 视频演示\n\u003Cdiv align=\"center\">\n  \u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fea4a7a8d-5d9e-4ccf-96e7-02f93797116e\" width=\"100%\" poster=\"\"> \u003C\u002Fvideo>\n\u003C\u002Fdiv>\n\n## 🔥 新闻\n- 2026 年 3 月 2 日：🎉 我们发布了 **LingBot-World-Base (Act)** 模型权重。\n- 2026 年 1 月 29 日：🎉 我们发布了 LingBot-World 的技术报告、代码和模型。\n\n\u003C!-- ## 🔖 LingBot-World 简介\n我们介绍了 **LingBot-World**，这是一个源自视频生成的**开源**世界模拟器。作为顶级世界模型，LingBot-World 提供以下功能。 \n- 它在广泛的环境（包括写实、科学场景、卡通风格等）中保持高保真度和稳健的动态效果。 \n- 它实现了分钟级的时间跨度，同时保持上下文随时间的一致性，这也被称为**长期记忆**。 \n- 它支持实时交互，在生成每秒 16 帧时延迟低于 1 秒。我们公开提供了代码和模型的访问权限，旨在缩小开源与闭源技术之间的差距。我们相信我们的发布将为社区赋能，使其在内容创作、游戏和机器人学习等领域获得实际应用。 -->\n\n## ⚙️ 快速开始\n此代码库基于 [Wan2.2](https:\u002F\u002Fgithub.com\u002FWan-Video\u002FWan2.2) 构建。安装说明请参阅其文档。\n### 安装\n克隆仓库：\n```sh\ngit clone https:\u002F\u002Fgithub.com\u002Frobbyant\u002Flingbot-world.git\ncd lingbot-world\n```\n安装依赖：\n```sh\n# Ensure torch >= 2.4.0\npip install -r requirements.txt\n```\n安装 [`flash_attn`](https:\u002F\u002Fgithub.com\u002FDao-AILab\u002Fflash-attention)：\n```sh\npip install flash-attn --no-build-isolation\n```\n### 模型下载\n\n| 模型 | 控制信号 | 分辨率 | 下载链接 |\n| :---  | :--- | :--- | :--- |\n| **LingBot-World-Base (Cam)** | 相机位姿 | 480P & 720P | 🤗 [HuggingFace](https:\u002F\u002Fhuggingface.co\u002Frobbyant\u002Flingbot-world-base-cam) 🤖 [ModelScope](https:\u002F\u002Fwww.modelscope.cn\u002Fmodels\u002FRobbyant\u002Flingbot-world-base-cam) |\n| **LingBot-World-Base (Act)** | 动作 | 480P & 720P | 🤗 [HuggingFace](https:\u002F\u002Fhuggingface.co\u002Frobbyant\u002Flingbot-world-base-act) |\n| **LingBot-World-Fast**       |    -    | - | *即将发布* |\n\n使用 huggingface-cli 下载模型：\n```sh\npip install \"huggingface_hub[cli]\"\nhuggingface-cli download robbyant\u002Flingbot-world-base-cam --local-dir .\u002Flingbot-world-base-cam\n```\n使用 modelscope-cli 下载模型：\n ```sh\npip install modelscope\nmodelscope download robbyant\u002Flingbot-world-base-cam --local_dir .\u002Flingbot-world-base-cam\n```\n\n### 推理\n在运行推理之前，您需要准备：\n- 输入图像\n- 文本提示\n- 控制信号（可选，可使用 [ViPE](https:\u002F\u002Fgithub.com\u002Fnv-tlabs\u002Fvipe) 从视频生成）\n  - `intrinsics.npy`（内参）：形状 `[num_frames, 4]`，其中 4 个值代表 `[fx, fy, cx, cy]`\n  - `poses.npy`（位姿）：形状 `[num_frames, 4, 4]`，其中每个 `[4, 4]` 代表 OpenCV 坐标系下的变换矩阵\n\n我们提供以下参考推理脚本：\n- `LingBot-World-Base (Cam)`：\n  - 480P：\n  ``` sh\n  torchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 480*832 --ckpt_dir lingbot-world-base-cam --image examples\u002F00\u002Fimage.jpg --action_path examples\u002F00 --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt \"The video presents a soaring journey through a fantasy jungle. The wind whips past the rider's blue hands gripping the reins, causing the leather straps to vibrate. The ancient gothic castle approaches steadily, its stone details becoming clearer against the backdrop of floating islands and distant waterfalls.\"\n  ```\n  - 720P：\n  ``` sh\n  torchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 720*1280 --ckpt_dir lingbot-world-base-cam --image examples\u002F00\u002Fimage.jpg --action_path examples\u002F00 --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt \"The video presents a soaring journey through a fantasy jungle. The wind whips past the rider's blue hands gripping the reins, causing the leather straps to vibrate. The ancient gothic castle approaches steadily, its stone details becoming clearer against the backdrop of floating islands and distant waterfalls.\"\n  ```\n  或者，您也可以在不使用控制信号的情况下运行推理：\n  ``` sh\n  torchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 480*832 --ckpt_dir lingbot-world-base-cam --image examples\u002F00\u002Fimage.jpg --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt \"The video presents a soaring journey through a fantasy jungle. The wind whips past the rider's blue hands gripping the reins, causing the leather straps to vibrate. The ancient gothic castle approaches steadily, its stone details becoming clearer against the backdrop of floating islands and distant waterfalls.\"\n  ```\n- `LingBot-World-Base (Act)`：\n  - 480P：\n  ``` sh\n  torchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 480*832 --ckpt_dir lingbot-world-base-act --image examples\u002F00\u002Fimage.jpg --action_path examples\u002F00 --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt \"The video presents a soaring journey through a fantasy jungle. The wind whips past the rider's blue hands gripping the reins, causing the leather straps to vibrate. The ancient gothic castle approaches steadily, its stone details becoming clearer against the backdrop of floating islands and distant waterfalls.\"\n  ```\n  - 720P：\n  ``` sh\n  torchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 720*1280 --ckpt_dir lingbot-world-base-act --image examples\u002F00\u002Fimage.jpg --action_path examples\u002F00 --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt \"The video presents a soaring journey through a fantasy jungle. The wind whips past the rider's blue hands gripping the reins, causing the leather straps to vibrate. The ancient gothic castle approaches steadily, its stone details becoming clearer against the backdrop of floating islands and distant waterfalls.\"\n  ```\n提示：\n如果您有足够的 CUDA 显存，可以将 `frame_num` 参数增加到 961 等数值，以生成 16 FPS 的一分钟视频。否则，如果 CUDA 显存不足，您可以使用 ``--t5_cpu`` 来降低显存占用。\n\n### 针对有限 GPU 资源的量化模型\n我们要衷心感谢社区对 LingBot-World 的宝贵支持和贡献。对于 GPU 显存有限的用户，我们推荐使用 LingBot-World-Base (Cam) 的 **4-bit 量化版本**，这能显著降低 GPU 显存消耗，同时在推理时保持具有竞争力的视觉质量。\n\n👉 下载链接：https:\u002F\u002Fhuggingface.co\u002Fcahlen\u002Flingbot-world-base-cam-nf4\n\n> ⚠️ 注意：此量化模型仅用于**推理**。与全精度模型相比，可能会出现轻微的视觉保真度和时间一致性下降。\n\n### 演示结果\n我们提供了对比演示，其中相机参数由 [ViPE](https:\u002F\u002Fgithub.com\u002Fnv-tlabs\u002Fvipe) 从从 [Genie3](https:\u002F\u002Fdeepmind.google\u002Fblog\u002Fgenie-3-a-new-frontier-for-world-models\u002F) 下载的原始视频中估计得出：\n\u003Cdiv align=\"center\">\n  \u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Ffc95ee9e-e8a9-4f70-9aa2-9536c8365ccc\" width=\"100%\" poster=\"\"> \u003C\u002Fvideo>\n  \u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fbac89021-b394-4f68-a688-9a0b90e30241\" width=\"100%\" poster=\"\"> \u003C\u002Fvideo>\n\u003C\u002Fdiv>\n\n## 📚 相关项目\n- [HoloCine](https:\u002F\u002Fholo-cine.github.io\u002F)\n- [Ditto](https:\u002F\u002Fezioby.github.io\u002FDitto_page\u002F)\n- [WorldCanvas](https:\u002F\u002Fworldcanvas.github.io\u002F)\n- [RewardForcing](https:\u002F\u002Freward-forcing.github.io\u002F)\n- [CoDeF](https:\u002F\u002Fqiuyu96.github.io\u002FCoDeF\u002F)\n\n## 📜 许可证\n本项目采用 Apache 2.0 许可证授权。请参阅 [LICENSE 文件](LICENSE.txt) 获取完整文本，包括权利和限制的详细信息。\n\n## ✨ 致谢\n我们要向 Wan Team 开源其代码和模型表示诚挚的感谢。他们的贡献对本项目的开发至关重要。\n\n## 📖 引用\n如果您发现这项工作对您的研究有用，请引用我们的论文：\n\n```\n@article{lingbot-world,\n      title={Advancing Open-source World Models}, \n      author={Robbyant Team and Zelin Gao and Qiuyu Wang and Yanhong Zeng and Jiapeng Zhu and Ka Leong Cheng and Yixuan Li and Hanlin Wang and Yinghao Xu and Shuailei Ma and Yihang Chen and Jie Liu and Yansong Cheng and Yao Yao and Jiayi Zhu and Yihao Meng and Kecheng Zheng and Qingyan Bai and Jingye Chen and Zehong Shen and Yue Yu and Xing Zhu and Yujun Shen and Hao Ouyang},\n      journal={arXiv preprint arXiv:2601.20540},\n      year={2026}\n}\n```","# LingBot-World 快速上手指南\n\n**LingBot-World** 是一个开源的世界模拟器，基于视频生成技术构建。它支持高保真环境、长时序记忆以及实时交互，适用于内容创作、游戏及机器人学习等领域。\n\n## 环境准备\n\n在开始之前，请确保您的开发环境满足以下要求：\n\n- **系统**: Linux \u002F Windows (推荐 Linux)\n- **Python**: 3.10+\n- **CUDA**: 兼容的 NVIDIA 显卡驱动\n- **PyTorch**: 版本需 >= 2.4.0\n- **依赖**: 本项目基于 [Wan2.2](https:\u002F\u002Fgithub.com\u002FWan-Video\u002FWan2.2)，部分依赖可参考其文档。\n\n## 安装步骤\n\n### 1. 克隆代码库\n\n```sh\ngit clone https:\u002F\u002Fgithub.com\u002Frobbyant\u002Flingbot-world.git\ncd lingbot-world\n```\n\n### 2. 安装依赖\n\n```sh\n# 确保 torch >= 2.4.0\npip install -r requirements.txt\n```\n\n### 3. 安装 Flash Attention\n\n```sh\npip install flash-attn --no-build-isolation\n```\n\n### 4. 下载模型权重\n\n推荐使用国内镜像源（ModelScope）以加速下载。\n\n**下载 Base (Cam) 模型：**\n\n```sh\npip install modelscope\nmodelscope download robbyant\u002Flingbot-world-base-cam --local_dir .\u002Flingbot-world-base-cam\n```\n\n*(注：如需下载 Act 模型，请将链接替换为 `robbyant\u002Flingbot-world-base-act`)*\n\n## 基本使用\n\n运行推理前，请准备好输入图片（如 `examples\u002F00\u002Fimage.jpg`）。\n\n### 基础推理命令\n\n以下示例展示了无控制信号的基础图像转视频生成（480P 分辨率）：\n\n```sh\ntorchrun --nproc_per_node=8 generate.py --task i2v-A14B --size 480*832 --ckpt_dir lingbot-world-base-cam --image examples\u002F00\u002Fimage.jpg --dit_fsdp --t5_fsdp --ulysses_size 8 --frame_num 161 --prompt \"The video presents a soaring journey through a fantasy jungle. The wind whips past the rider's blue hands gripping the reins, causing the leather straps to vibrate. The ancient gothic castle approaches steadily, its stone details becoming clearer against the backdrop of floating islands and distant waterfalls.\"\n```\n\n### 显存优化建议\n\n- **低显存用户**：如果 CUDA 显存不足，可添加 `--t5_cpu` 参数以降低内存占用。\n- **长视频生成**：若显存充足，可将 `frame_num` 调整为更大值（如 961）以生成约 1 分钟的视频（16 FPS）。\n- **量化模型**：对于资源有限的用户，推荐使用 4-bit 量化版本（[HuggingFace 链接](https:\u002F\u002Fhuggingface.co\u002Fcahlen\u002Flingbot-world-base-cam-nf4)），但仅用于推理，画质可能略有下降。","某自动驾驶研究团队正在开发城市路况下的避障算法，急需构建大规模、多样化的虚拟测试场景来替代昂贵的实车路测。\n\n### 没有 lingbot-world 时\n- 依赖传统游戏引擎手动建模，创建复杂街道和天气效果耗时数周，效率极低。\n- 生成的测试视频缺乏物理连贯性，车辆行驶轨迹容易出现逻辑断裂或穿模现象。\n- 无法进行实时交互反馈，调整传感器参数后必须重新跑完整个长序列才能看到结果。\n- 高质量闭源仿真模型授权费用高昂，限制了团队对边缘案例（Corner Case）的探索广度。\n\n### 使用 lingbot-world 后\n- lingbot-world 基于视频生成技术自动构建高保真环境，支持多种风格与科学场景，建模时间缩短至分钟级。\n- 具备分钟级长程记忆能力，确保车辆在长时间行驶中环境光照、物体位置保持一致，符合现实逻辑。\n- 提供低于 1 秒延迟的实时交互接口，研究人员可即时输入控制指令并观察 16 帧\u002F秒的动态反馈。\n- 完全开源免费且支持动作控制信号，团队能低成本快速迭代数千种极端驾驶场景以优化算法鲁棒性。\n\nlingbot-world 通过开放的高保真动态世界模拟能力，将机器人算法的训练周期从数月压缩至数周，显著降低了研发风险与成本。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FRobbyant_lingbot-world_f7bb75f7.png","Robbyant","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FRobbyant_fb160f19.png","Intelligence in Action, Benefits for Everyone",null,"robbyant_brain","https:\u002F\u002Fwww.robbyant.com\u002F","https:\u002F\u002Fgithub.com\u002FRobbyant",[83],{"name":84,"color":85,"percentage":86},"Python","#3572A5",100,3319,273,"2026-04-05T10:53:53","Apache-2.0","未说明","需要 NVIDIA GPU（Flash Attention 依赖），示例配置为 8 卡并行，显存需求较高，低显存用户可使用 4-bit 量化版",{"notes":94,"python":91,"dependencies":95},"代码基于 Wan2.2 构建；推理命令默认使用 torchrun 进行 8 卡分布式部署；需准备输入图像、文本提示及可选的相机位姿控制信号（intrinsics.npy, poses.npy）；若显存不足可添加 --t5_cpu 参数或使用量化模型",[96,97,98,99],"torch>=2.4.0","flash-attn","huggingface_hub","modelscope",[14,35],[102,103,104,67,105],"image-to-video","video-generation","world-models","aigc",9,"2026-03-27T02:49:30.150509","2026-04-06T06:53:12.722816",[110,115,120,125,130,135,140,145],{"id":111,"question_zh":112,"answer_zh":113,"source_url":114},2728,"运行项目需要多少 GPU 显存？","根据社区测试反馈，运行 480p 模型在 4 张 H100 (80GB) 显卡上时，峰值总 VRAM 使用量约为 196GiB。如果显存不足，可能需要限制并行任务数量以避免内存溢出。","https:\u002F\u002Fgithub.com\u002FRobbyant\u002Flingbot-world\u002Fissues\u002F1",{"id":116,"question_zh":117,"answer_zh":118,"source_url":119},2729,"Flash-Attention 的版本有什么要求？","Flash-Attention 对 CUDA 和 Python 版本比较挑剔。建议搭配 PyTorch stable 2.10.0 + CUDA 12.8 + Python 3.11 进行编译（尽量避免使用 nightly 版本）。编译过程可能需要数小时，完成后会生成较大的 `.pyd` 文件。","https:\u002F\u002Fgithub.com\u002FRobbyant\u002Flingbot-world\u002Fissues\u002F7",{"id":121,"question_zh":122,"answer_zh":123,"source_url":124},2730,"如何测试“可提示的世界事件”功能？","对于已发布的双向模型，只需在文本提示词中直接注入目标世界事件即可。除了提供初始帧和特定轨迹\u002F动作外，可以在基础场景描述中加入如“夜空中绽放烟花”、“一条龙飞入画面”等短语来触发事件。","https:\u002F\u002Fgithub.com\u002FRobbyant\u002Flingbot-world\u002Fissues\u002F34",{"id":126,"question_zh":127,"answer_zh":128,"source_url":129},2731,"如何使用自己的数据准备 intrinsics.npy 和 poses.npy？","可以使用 [ViPE](https:\u002F\u002Fgithub.com\u002Fnv-tlabs\u002Fvipe) 工具从参考视频中提取相机内参（intrinsics）和外参（extrinsics）。具体操作请参考项目 README 中的推理（Inference）部分演示。","https:\u002F\u002Fgithub.com\u002FRobbyant\u002Flingbot-world\u002Fissues\u002F2",{"id":131,"question_zh":132,"answer_zh":133,"source_url":134},2732,"ViPE 生成的控制信号文件格式不匹配怎么办？","ViPE 输出的 `.npz` 文件和 LingBot 需要的 `.npy` 文件存储的是相同信息。官方建议编写脚本来进行格式转换，利用视频编码库可以较容易地完成这一过程。","https:\u002F\u002Fgithub.com\u002FRobbyant\u002Flingbot-world\u002Fissues\u002F22",{"id":136,"question_zh":137,"answer_zh":138,"source_url":139},2733,"多卡运行时报错 SIGSEGV 或显存溢出如何处理？","如果遇到进程终止信号 11 (SIGSEGV)，请检查 PyTorch 环境是否正常。此外，可以通过设置 `--t5_cpu` 参数将 T5 模型卸载到 CPU，从而进一步降低显存占用并避免崩溃。","https:\u002F\u002Fgithub.com\u002FRobbyant\u002Flingbot-world\u002Fissues\u002F21",{"id":141,"question_zh":142,"answer_zh":143,"source_url":144},2734,"如何在低配显卡（如单卡 RTX 系列）上运行？","可以尝试使用社区发布的轻量化 checkpoint。相关资源地址为：https:\u002F\u002Fhuggingface.co\u002Frobbyant\u002Flingbot-world-base-cam\u002Ftree\u002Fmain。","https:\u002F\u002Fgithub.com\u002FRobbyant\u002Flingbot-world\u002Fissues\u002F15",{"id":146,"question_zh":147,"answer_zh":148,"source_url":149},2735,"WSL 环境下无法识别 PyTorch 如何解决？","经排查，该问题通常是因为 CUDA 环境问题。建议在 Linux 子系统下重新安装 CUDA 驱动以确保环境兼容性。","https:\u002F\u002Fgithub.com\u002FRobbyant\u002Flingbot-world\u002Fissues\u002F23",[]]