[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-intel--AI-Playground":3,"tool-intel--AI-Playground":64},[4,17,27,35,43,56],{"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},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"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,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"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",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"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",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":70,"readme_en":71,"readme_zh":72,"quickstart_zh":73,"use_case_zh":74,"hero_image_url":75,"owner_login":76,"owner_name":77,"owner_avatar_url":78,"owner_bio":79,"owner_company":80,"owner_location":80,"owner_email":81,"owner_twitter":80,"owner_website":80,"owner_url":82,"languages":83,"stars":114,"forks":115,"last_commit_at":116,"license":117,"difficulty_score":23,"env_os":118,"env_gpu":119,"env_ram":120,"env_deps":121,"category_tags":130,"github_topics":80,"view_count":23,"oss_zip_url":80,"oss_zip_packed_at":80,"status":16,"created_at":131,"updated_at":132,"faqs":133,"releases":163},2712,"intel\u002FAI-Playground","AI-Playground","AI PC starter app for doing AI image creation, image stylizing, and chatbot on a PC powered by an Intel® Arc™ GPU.","AI-Playground 是一款专为搭载英特尔® Arc™显卡的 AI PC 设计的开源本地应用套件，旨在让用户在完全离线的环境下，轻松体验聊天对话、代码辅助、文档检索、图像分析及音视频生成等全套生成式 AI 功能。\n\n它主要解决了用户依赖云端服务时面临的数据隐私泄露风险、网络延迟及订阅费用高昂等痛点。作为 Gemini、ChatGPT 等云服务的本地替代方案，AI-Playground 无需将敏感个人信息上传至第三方服务器，且完全免费。用户无需具备复杂的后端框架配置知识，即可通过单一界面调用多种先进模型。\n\n这款软件非常适合注重数据隐私的普通消费者、对 AI 充满好奇的进阶用户，以及希望快速原型验证的设计师或开发者。其独特技术亮点在于深度优化了英特尔硬件性能，支持包括 Qwen3 VL、DeepSeek R1、Llama 3.2 在内的最新大语言模型，以及 Stable Diffusion、Flux.1 等主流图像生成模型。它不仅提供强大的视觉推理与文档问答（RAG）能力，还集成了无订阅费的图像编辑功能，如超分辨率、局部重绘及 2D 转 3D 网格生成，让创意工作流更加安全、高效且","AI-Playground 是一款专为搭载英特尔® Arc™显卡的 AI PC 设计的开源本地应用套件，旨在让用户在完全离线的环境下，轻松体验聊天对话、代码辅助、文档检索、图像分析及音视频生成等全套生成式 AI 功能。\n\n它主要解决了用户依赖云端服务时面临的数据隐私泄露风险、网络延迟及订阅费用高昂等痛点。作为 Gemini、ChatGPT 等云服务的本地替代方案，AI-Playground 无需将敏感个人信息上传至第三方服务器，且完全免费。用户无需具备复杂的后端框架配置知识，即可通过单一界面调用多种先进模型。\n\n这款软件非常适合注重数据隐私的普通消费者、对 AI 充满好奇的进阶用户，以及希望快速原型验证的设计师或开发者。其独特技术亮点在于深度优化了英特尔硬件性能，支持包括 Qwen3 VL、DeepSeek R1、Llama 3.2 在内的最新大语言模型，以及 Stable Diffusion、Flux.1 等主流图像生成模型。它不仅提供强大的视觉推理与文档问答（RAG）能力，还集成了无订阅费的图像编辑功能，如超分辨率、局部重绘及 2D 转 3D 网格生成，让创意工作流更加安全、高效且可控。","\u003Ca href=\"https:\u002F\u002Fscan.coverity.com\u002Fprojects\u002Fai-playground\">\n  \u003Cimg alt=\"Coverity Scan Build Status\"\n       src=\"https:\u002F\u002Fscan.coverity.com\u002Fprojects\u002F30694\u002Fbadge.svg\"\u002F>\n\u003C\u002Fa>\n\n# AI PLAYGROUND 3.0.3 beta\n\u003Cimg width=\"2025\" height=\"593\" alt=\"image\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fintel_AI-Playground_readme_d561605aac06.png\" \u002F>\n\n\nWelcome to AI Playground open source project and AI PC generative AI application suite. This application provides a full suite of generative AI features for chat, code assistance, document search, image analysis, image and video generation. All features run offline and are powered by your PC’s Intel® Core™ Ultra with built-in Intel Arc GPU or Intel Arc™ dGPU Series A or B with 8GB+ of vRAM.\n\nAI Playground is intended to act as an offline alternative to cloud tools such Gemini ChatGPT and Grok.  AI Playground leverages libraries from GitHub and Huggingface including:\n- Image Diffusion (PyTorch 2.10): Stable Diffusion 1.5, SDXL, Flux.1-Schnell, Flux.1 Kontext[dev], Z-Image, Wan2.1 VACE, LTX-Video\n- LLM: GGUF (Llama.cpp Vulkan) - Qwen3 VL, GPT-OSS 20B, DeepSeek R1 Distilled, Phi3, Mistral 7B, Llama 3.2: OpenVINO - TinyLlama, Mistral 7B, Phi3 mini, Phi3.5 mini, DeepSeek R1 Distill (1.5B, 7B)\n\u003Cimg width=\"3221\" height=\"1849\" alt=\"image\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fintel_AI-Playground_readme_25487a23570e.png\" \u002F>\n\nAs a local alternative to cloud AI service, AI Playground is intended to give consumers and AI curious prosumers easy and intuitive access to a wide variety of generative AI features using their Intel powered AI PC. This means you can be offline, without loading sensitive or personal data to 3rd party sites, for free, in a single app without having to know how to install and manage multiple AI backend frameworks.   Key features:\n- Latest and greatest chat models: Support for Qwen 3 VL, Mistral 7B, DeepSeek R1 or GPT-OSS, AI playground makes a variety of chat models available to users\n- Vision, Reasoning and RAG: Chat features support Vision, Reasoning and RAG to analyze and get deep answers on both visual and text content\nAnalyze images with Qwen3 VL Model\tVibe Coding with GPT-OSS 20B Reasoning\tDocument RAG with Mistral 7B Instruct\n- Image Generation: From Stable Diffusion 1.5, SDXL, Flux.1 and Z-image models AI Playground is making a breadth of image generation from quick easy low-res draft generation to high quality image generation\n- Image Editing: Subscription free and private control for upscaling, inpainting, outpainting, 2D to 3D mesh or editing images in a variety of ways.  Good for editing personal photos to taking sketches and generated images to the next level with greater control. \n\n## README.md\n- English (readme.md)\n\n## Min Specs\nAI Playground alpha and beta installers are currently available downloadable executables, or available as a source code from our Github repository.  To run AI Playground you must have a PC that meets the following specifications\n\n*\tWindows OS\n*\tIntel Core Ultra Series 3, Series 2H, Series 2V, or Series 1 H processor OR Intel Arc GPU Series A or Series B (discrete) with 8GB of vRAM\n\n## Installation - Packaged Installer: \nThis is a single packaged installer for all supported hardware mentioned above. This installer simplifies the process for end users to install AI Playground on their PCs. Please note that while this makes the installation process easier, this is open-source beta software, and there may be component and version conflicts. Refer to the Troubleshooting section for known issues.\n\n### Download the installer\n:new: **AI Playground 3.0.2 beta (all SKUs)** - [Release Notes](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Ftag\u002Fv3.0.3-beta_hf) | [Download](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Fdownload\u002Fv3.0.3-beta_hf\u002FAI-Playground-installer.exe) :new:\n\n### Installation Process for v3.0\n1. The installer only installs the Electron frontend, so it completes very quickly.\n2. On the first run, AI Playground Setup window appears where you install needed backend components for AI Playground to function properly. This process requires a strong and open network and may **take several minutes**.\n3. Download the Users Guide for application information: [AI Playground Users Guide](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Fblob\u002Fmain\u002FAI%20Playground%20Users%20Guide.pdf)\n\n### Troubleshooting Installation\nThe following are known situations where your installation may be blocked or interrupted.  Review the following to remedy installations issues.  If installation issues persist, generate a copy of the log by typing CTRL+SHIFT+I, select the console tab and copy the last few entries of the log written where the installer failed.  Provide these details to us via the issues tab here, or via the Intel Insiders Discord, or Graphics forum on Intel's support site.\n\n1. **Llama.cpp embedding issues**: At the time of this release, Llama.cpp embeddings may have issues with:\n  * Recent drivers, and may require DDU to clean driver cache.\n  * Anti-Virus software - features needed to read and write embedding cache may not be properly installed:  Disable anti-virus, restart \n2. **Restart**: Time-out issues have been sighted, which show as a failed install but resolve when restarting AI Playground\n3. **Verify Intel Arc GPU**: Ensure your system has an Intel Arc GPU with the lastest driver. Go to your Windows Start Menu, type \"Device Manager,\" and under Display Adapters, check the name of your GPU device. It should describe an Intel Arc GPU. If so, then you you have a GPU that means our minimum specifications.  If it says \"Intel(R) Graphics,\" your system does not have a built-in Intel Arc GPU and does not meet the minimum specifications. If your GPU is an discrete GPU such as Intel Arc A or B series GPU, then you can troubleshoot a troubled installation by disabling the iGPU in Device Manager\n4. **Interrupted Installation**: The online installation for backend components can be interrupted or blocked by an IT network, firewall, or sleep settings. Ensure you are on an open network, with the firewall off, and set sleep settings to stay awake when powered on.\n5. **Missing Libraries**: Some Windows systems may be missing needed libraries. This can be fixed by installing the 64-bit VC++ redistribution from Microsoft [here](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fcpp\u002Fwindows\u002Flatest-supported-vc-redist?view=msvc-170). It is recommended this be done after updating the Graphics drivers. Then install AI Playground.\n6. **Python Conflict**: Some PCs with an existing installation of Python can cause a conflict with AI Playground installation, where the wrong or conflicting packages are installed due to the incorrect version or location of Python on the system.  This is usually remedied by uninstalling Python environment, restarting and reinstalling AI Playground\n7.  **Temp Files**: Should the installation be interrupted because of any of the above issues it is possible that temporary installation files have been left behind and trying to install with these files in place can block the installation. Remove these files or do a clean install of AI Playground to remedy\n\n## Project Development\n### Checkout Source Code\n\nTo get started, clone the repository and navigate to the project directory:\n\n```cmd\ngit clone -b dev https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground.git\ncd AI-Playground\n```\n\n### Install Node.js Dependencies\n\n1. Install the Node.js development environment from [Node.js](https:\u002F\u002Fnodejs.org\u002Fen\u002Fdownload).\n\n2. Navigate to the `WebUI` directory and install all Node.js dependencies:\n\n```cmd\ncd WebUI\nnpm install\n```\n\n### Fetch External Resources\n\n1. In the `WebUI` directory, execute the `fetch-external-resources` script to download required external resources:\n\nThis will download `uv` (Python package manager) and other required tools to the `build\u002Fresources\u002F` directory.\n\n### Launch the application\n\nTo start the application in development mode, run:\n\n```\nnpm run dev\n```\n\n### (Optional) Build the installer\n\nTo build the installer, run:\n\n```\nnpm run build\n```\n\nThe installer executable will be located in the `build\u002Felectron` folder.\n\n## Model Support\nAI Playground does not ship with any generative AI models but does make models available for all features either directly from the interface or indirectly by the users downloading models from HuggingFace.co or CivitAI.com and placing them in the appropriate model folder. \n\nModels currently linked from the application \n\n### AI Model & License Registry\n\n| Model Path \u002F Name | Model Card (HF) | License Link |\n| :--- | :--- | :--- |\n| AdamCodd\u002Fvit-base-nsfw-detector | [Model Card](https:\u002F\u002Fhuggingface.co\u002FAdamCodd\u002Fvit-base-nsfw-detector) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Aitrepreneur\u002Finsightface\u002Finswapper_128.onnx | [Model Card](https:\u002F\u002Fhuggingface.co\u002FAitrepreneur\u002Finsightface) | [Non-Commercial](https:\u002F\u002Fhuggingface.co\u002FAitrepreneur\u002Finsightface#license) |\n| alimama-creative\u002FFLUX.1-Turbo-Alpha | [Model Card](https:\u002F\u002Fhuggingface.co\u002Falimama-creative\u002FFLUX.1-Turbo-Alpha) | [FLUX.1-dev License](https:\u002F\u002Fhuggingface.co\u002Fblack-forest-labs\u002FFLUX.1-dev\u002Fblob\u002Fmain\u002FLICENSE.md) |\n| BGE Small EN v1.5 (GGUF) | [Model Card](https:\u002F\u002Fhuggingface.co\u002FBAAI\u002Fbge-small-en-v1.5) | [MIT License](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT) |\n| black-forest-labs\u002FFLUX.2-klein-4b-fp8\u002Fflux-2-klein-4b-fp8.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fblack-forest-labs\u002FFLUX.2-klein-4b-fp8) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| city96\u002Ft5-v1_1-xxl-encoder-gguf\u002Ft5-v1_1-xxl-encoder-Q3_K_M.gguf | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fcity96\u002Ft5-v1_1-xxl-encoder-gguf) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| city96\u002Ft5-v1_1-xxl-encoder-gguf\u002Ft5-v1_1-xxl-encoder-Q4_K_M.gguf | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fcity96\u002Ft5-v1_1-xxl-encoder-gguf) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| city96\u002Fumt5-xxl-encoder-gguf\u002Fumt5-xxl-encoder-Q4_K_M.gguf | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fcity96\u002Fumt5-xxl-encoder-gguf) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| comfyanonymous\u002Fflux_text_encoders\u002Fclip_l.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fcomfyanonymous\u002Fflux_text_encoders) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| comfyanonymous\u002Fflux_text_encoders\u002Ft5xxl_fp8_e4m3fn_scaled.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fcomfyanonymous\u002Fflux_text_encoders) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Comfy-Org\u002Fflux1-kontext-dev\u002Fflux1-dev-kontext_fp8_scaled.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002Fflux1-kontext-dev) | [FLUX.1-dev License](https:\u002F\u002Fhuggingface.co\u002Fblack-forest-labs\u002FFLUX.1-dev\u002Fblob\u002Fmain\u002FLICENSE.md) |\n| Comfy-Org\u002FLumina_Image_2.0_Repackaged\u002Fae.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002FLumina_Image_2.0_Repackaged) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Comfy-Org\u002FReal-ESRGAN_repackaged\u002FRealESRGAN_x4plus.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002FReal-ESRGAN_repackaged) | [BSD-3-Clause](https:\u002F\u002Fopensource.org\u002Flicenses\u002FBSD-3-Clause) |\n| Comfy-Org\u002FWan_2.1_ComfyUI_repackaged\u002Fwan_2.1_vae.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002FWan_2.1_ComfyUI_repackaged) | [Wan 2.1 License](https:\u002F\u002Fhuggingface.co\u002FWan-AI\u002FWan2.1-T2V-14B\u002Fblob\u002Fmain\u002FLICENSE) |\n| Comfy-Org\u002Fz_image_turbo\u002Fae.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002Fz_image_turbo) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Comfy-Org\u002Fz_image_turbo\u002Fqwen_3_4b.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002Fz_image_turbo) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Comfy-Org\u002Fz_image_turbo\u002Fz_image_turbo_bf16.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002Fz_image_turbo) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| DeepSeek-R1-Distill-Qwen 1.5B | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fdeepseek-ai\u002FDeepSeek-R1-Distill-Qwen-1.5B) | [MIT License](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT) |\n| DeepSeek-R1-Distill-Qwen 7B | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fdeepseek-ai\u002FDeepSeek-R1-Distill-Qwen-7B) | [MIT License](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT) |\n| Gemma 3 4B IT (Unsloth) | [Model Card](https:\u002F\u002Fhuggingface.co\u002Funsloth\u002Fgemma-3-4b-it) | [Gemma License](https:\u002F\u002Fai.google.dev\u002Fgemma\u002Fterms) |\n| gmk123\u002FGFPGAN\u002FGFPGANv1.4.pth | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fgmk123\u002FGFPGAN) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| GPT-OSS 20B (Unsloth) | [Model Card](https:\u002F\u002Fhuggingface.co\u002Funsloth\u002Fgpt-oss-20b) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| InternVL2 4B (OV) | [Model Card](https:\u002F\u002Fhuggingface.co\u002FOpenGVLab\u002FInternVL2-4B) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| latent-consistency\u002Flcm-lora-sdv1-5\u002Fpytorch_lora_weights.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002Flatent-consistency\u002Flcm-lora-sdv1-5) | [OpenRAIL++](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| latent-consistency\u002Flcm-lora-sdxl\u002Fpytorch_lora_weights.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002Flatent-consistency\u002Flcm-lora-sdxl) | [OpenRAIL++](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Lightricks\u002FLTX-Video\u002Fltxv-2b-0.9.6-distilled-04-25.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-Video) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Llama 3.2 3B Instruct | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FLlama-3.2-3B-Instruct) | [Llama 3.2 License](https:\u002F\u002Fgithub.com\u002Fmeta-llama\u002Fllama-models\u002Fblob\u002Fmain\u002Fmodels\u002Fllama3_2\u002FLICENSE) |\n| lllyasviel\u002Ffooocus_inpaint\u002Ffooocus_inpaint_head.pth | [Model Card](https:\u002F\u002Fhuggingface.co\u002Flllyasviel\u002Ffooocus_inpaint) | [OpenRAIL](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| lllyasviel\u002Ffooocus_inpaint\u002Finpaint_v26.fooocus.patch | [Model Card](https:\u002F\u002Fhuggingface.co\u002Flllyasviel\u002Ffooocus_inpaint) | [OpenRAIL](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Lykon\u002FDreamShaper\u002FDreamShaper_8_pruned.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FLykon\u002FDreamShaper) | [OpenRAIL-M](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Lykon\u002Fdreamshaper-8-inpainting\u002Ftext_encoder\u002Fmodel.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FLykon\u002Fdreamshaper-8-inpainting) | [OpenRAIL-M](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Lykon\u002Fdreamshaper-8-inpainting\u002Funet\u002Fmodel.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FLykon\u002Fdreamshaper-8-inpainting) | [OpenRAIL-M](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Lykon\u002Fdreamshaper-8-inpainting\u002Fvae\u002Fmodel.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FLykon\u002Fdreamshaper-8-inpainting) | [OpenRAIL-M](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Meta-Llama 3.1 8B Instruct | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FMeta-Llama-3.1-8B-Instruct) | [Llama 3.1 License](https:\u002F\u002Fgithub.com\u002Fmeta-llama\u002Fllama-models\u002Fblob\u002Fmain\u002Fmodels\u002Fllama3_1\u002FLICENSE) |\n| Mistral 7B Instruct v0.2 (OV) | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FMistral-7B-Instruct-v0.2) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Mistral 7B Instruct v0.3 | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FMistral-7B-Instruct-v0.3) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Mistral 7B Instruct v0.3 (OV) | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FMistral-7B-Instruct-v0.3) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Nomic Embed Text v1.5 (GGUF) | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fnomic-ai\u002Fnomic-embed-text-v1.5) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Phi-3 Mini 4k Instruct (OV) | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fmicrosoft\u002FPhi-3-mini-4k-instruct) | [MIT License](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT) |\n| Phi-3.5 Mini Instruct (OV) | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fmicrosoft\u002FPhi-3.5-mini-instruct) | [MIT License](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT) |\n| QuantStack\u002FWan2.1_14B_VACE-GGUF\u002FWan2.1_14B_VACE-Q8_0.gguf | [Model Card](https:\u002F\u002Fhuggingface.co\u002FQuantStack\u002FWan2.1_14B_VACE-GGUF) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Qwen2-VL 7B Instruct (OV) | [Model Card](https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen2-VL-7B-Instruct) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Qwen3 4B (OV) | [Model Card](https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen2.5-4B) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Qwen3 4B (Unsloth) | [Model Card](https:\u002F\u002Fhuggingface.co\u002Funsloth\u002FQwen2.5-4B) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Qwen3 4B Instruct 2507 (Unsloth) | [Model Card](https:\u002F\u002Fhuggingface.co\u002Funsloth\u002FQwen2.5-4B-Instruct) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Qwen3-VL 4B Instruct (Unsloth) | [Model Card](https:\u002F\u002Fhuggingface.co\u002Funsloth\u002FQwen2-VL-7B-Instruct) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| RunDiffusion\u002FJuggernaut-XL-v9\u002FRunDiffusionPhoto_v2.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002FRunDiffusion\u002FJuggernaut-XL-v9) | [OpenRAIL-M](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| SmolLM2 1.7B Instruct | [Model Card](https:\u002F\u002Fhuggingface.co\u002FHuggingFaceTB\u002Fsmollm2-1.7b-instruct) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| stabilityai\u002Fcontrol-lora\u002Frank128-canny-rank128.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fstabilityai\u002Fcontrol-lora) | [SAI Community](https:\u002F\u002Fhuggingface.co\u002Fstabilityai\u002Fcontrol-lora#license) |\n| tencent\u002FHunyuan3D-2.1\u002Fhunyuan3d-dit-v2-1\u002Fmodel.fp16.ckpt | [Model Card](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuan3D-2.1) | [Hunyuan3D License](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuan3D-2.1\u002Fblob\u002Fmain\u002FLICENSE.txt) |\n| tencent\u002FHunyuan3D-2\u002Fhunyuan3d-dit-v2-0\u002Fmodel.fp16.safetensors | [Model Card](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuan3D-2) | [Hunyuan3D License](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuan3D-2\u002Fblob\u002Fmain\u002FLICENSE.txt) |\n| TinyLlama 1.1B Chat (OV) | [Model Card](https:\u002F\u002Fhuggingface.co\u002FTinyLlama\u002FTinyLlama-1.1B-Chat-v1.0) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Whisper (OV) | [Model Card](https:\u002F\u002Fhuggingface.co\u002Fopenai\u002Fwhisper-large-v3) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n\n\n\nBe sure to check license terms for any model used in AI Playground especially taking note of any restrictions.\n\n### Use Alternative Models\nCheck the [User Guide](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground\u002Fblob\u002Fmain\u002FAI%20Playground%20Users%20Guide.pdf) for details or [watch this video](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=1FXrk9Xcx2g) on how to add alternative Stable Diffusion models to AI Playground\n\n\n## Notices and Disclaimers: \nFor information on AI Playground terms, license and disclaimers, visit the project and files on GitHub repo:\u003C\u002Fbr >\n[License](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground\u002Fblob\u002Fmain\u002FLICENSE) | [Notices & Disclaimers](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground\u002Fblob\u002Fmain\u002Fnotices-disclaimers.md)\n\nThe software may include third party components with separate legal notices or governed by other agreements, as may be described in the Third Party Notices file accompanying the software.\n\n## Credit\nLicense details for borrowed code and components can be found in our [3rdpartynoticeslicense](3rdpartynoticeslicenses.txt) file.  \nAdditionally, these entities and their work stand out as are fundamental to AI Playground.\n*\tPyTorch - https:\u002F\u002Fpytorch.org\u002F \n*\tStable Diffusion - https:\u002F\u002Fgithub.com\u002FStability-AI\u002Fstablediffusion\n*\tComfyUI -  https:\u002F\u002Fgithub.com\u002Fcomfyanonymous\u002FComfyUI\n*\tOpenVINO - https:\u002F\u002Fopenvinotoolkit.github.io\u002Fopenvino.genai\u002F \n*\tLlama.cpp - https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp \n*\tVue.js - https:\u002F\u002Fvuejs.org\u002F \n*\tPlus countless other open-source projects and contributors that make this work possible!\n\n","\u003Ca href=\"https:\u002F\u002Fscan.coverity.com\u002Fprojects\u002Fai-playground\">\n  \u003Cimg alt=\"Coverity Scan Build Status\"\n       src=\"https:\u002F\u002Fscan.coverity.com\u002Fprojects\u002F30694\u002Fbadge.svg\"\u002F>\n\u003C\u002Fa>\n\n# AI 游乐场 3.0.3 测试版\n\u003Cimg width=\"2025\" height=\"593\" alt=\"图片\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fintel_AI-Playground_readme_d561605aac06.png\" \u002F>\n\n欢迎来到 AI 游乐场开源项目及 AI PC 生成式 AI 应用套件。本应用提供全套的生成式 AI 功能，包括聊天、代码辅助、文档搜索、图像分析以及图像和视频生成。所有功能均可离线运行，并由您电脑上的英特尔® Core™ Ultra 处理器（内置英特尔 Arc GPU）或英特尔 Arc™ dGPU A 或 B 系列显卡（配备 8GB 及以上显存）驱动。\n\nAI 游乐场旨在作为 Gemini、ChatGPT 和 Grok 等云端工具的离线替代方案。它利用了来自 GitHub 和 Huggingface 的多种库，包括：\n- 图像扩散模型（PyTorch 2.10）：Stable Diffusion 1.5、SDXL、Flux.1-Schnell、Flux.1 Kontext[dev]、Z-Image、Wan2.1 VACE、LTX-Video\n- 大语言模型：GGUF（Llama.cpp Vulkan）——Qwen3 VL、GPT-OSS 20B、DeepSeek R1 Distilled、Phi3、Mistral 7B、Llama 3.2；OpenVINO——TinyLlama、Mistral 7B、Phi3 mini、Phi3.5 mini、DeepSeek R1 Distill（1.5B、7B）\n\u003Cimg width=\"3221\" height=\"1849\" alt=\"图片\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fintel_AI-Playground_readme_25487a23570e.png\" \u002F>\n\n作为云端 AI 服务的本地替代方案，AI 游乐场旨在让普通消费者和对 AI 感兴趣的产消者能够通过搭载英特尔处理器的 AI PC，轻松直观地使用各种生成式 AI 功能。这意味着您可以在离线状态下，无需将敏感或个人数据上传至第三方平台，免费使用一款应用即可完成多项任务，而无需掌握如何安装和管理多个 AI 后端框架。主要特点如下：\n- 最新最强大的聊天模型：支持 Qwen 3 VL、Mistral 7B、DeepSeek R1 或 GPT-OSS 等模型，AI 游乐场为用户提供了丰富的聊天选择。\n- 视觉、推理与 RAG：聊天功能支持视觉理解、推理和 RAG 技术，可针对图像和文本内容进行深度分析并给出详尽答案。\n使用 Qwen3 VL 模型分析图像\t借助 GPT-OSS 20B 推理进行编程\t使用 Mistral 7B Instruct 进行文档 RAG\n- 图像生成：从 Stable Diffusion 1.5、SDXL、Flux.1 到 Z-image 等模型，AI 游乐场提供了从快速简单的低分辨率草图生成到高质量图像生成的广泛选项。\n- 图像编辑：无需订阅，即可私密地进行图像放大、局部修复、扩展绘制等操作，或将 2D 图像转换为 3D 模型等多种编辑方式。非常适合编辑个人照片，或进一步提升草图和生成图像的质量，获得更精细的控制。\n\n## README.md\n- 英文（readme.md）\n\n## 最低配置要求\n目前，AI 游乐场的 Alpha 和 Beta 版安装程序以可下载的可执行文件形式提供，也可从我们的 GitHub 仓库获取源代码。要运行 AI 游乐场，您的电脑必须满足以下规格：\n\n* Windows 操作系统\n* 英特尔 Core Ultra 系列 3、系列 2H、系列 2V 或系列 1H 处理器，或配备 8GB 显存的英特尔 Arc GPU A 或 B 系列独立显卡\n\n## 安装——打包安装程序：\n这是适用于上述所有支持硬件的单个打包安装程序。该安装程序简化了终端用户在电脑上安装 AI 游乐场的过程。请注意，尽管此安装程序使流程更加便捷，但本软件仍属开源测试版，可能存在组件和版本冲突问题。有关已知问题，请参阅故障排除部分。\n\n### 下载安装程序\n:new: **AI 游乐场 3.0.2 测试版（所有 SKU）** —— [发布说明](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Ftag\u002Fv3.0.3-beta_hf) | [下载](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Fdownload\u002Fv3.0.3-beta_hf\u002FAI-Playground-installer.exe) :new:\n\n### v3.0 版本安装流程\n1. 该安装程序仅安装 Electron 前端，因此安装过程非常迅速。\n2. 首次运行时，会弹出 AI 游乐场设置窗口，您需要在此处安装必要的后端组件，以确保 AI 游乐场正常运行。此过程需要稳定且开放的网络连接，可能**耗时数分钟**。\n3. 请下载用户指南以获取应用相关信息：[AI 游乐场用户指南](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Fblob\u002Fmain\u002FAI%20Playground%20Users%20Guide.pdf)\n\n### 安装故障排除\n以下是可能导致安装被阻止或中断的已知情况。请查看以下内容以解决安装问题。如果安装问题仍然存在，请按 CTRL+SHIFT+I 生成日志副本，选择“控制台”选项卡，并复制安装程序失败处写入的日志最后几条记录。请通过此处的问题标签页、Intel Insiders Discord 或 Intel 支持网站上的图形论坛向我们提供这些详细信息。\n\n1. **Llama.cpp 嵌入问题**：在本版本发布时，Llama.cpp 嵌入可能存在以下问题：\n  * 最新驱动程序可能存在问题，需要使用 DDU 清理驱动缓存。\n  - 防病毒软件：读取和写入嵌入缓存所需的某些功能可能未正确安装：请禁用防病毒软件并重启。\n2. **重启**：曾出现超时问题，表现为安装失败，但重启 AI Playground 后即可解决。\n3. **验证 Intel Arc GPU**：确保您的系统配备了最新驱动程序的 Intel Arc GPU。进入 Windows 开始菜单，输入“设备管理器”，在“显示适配器”下检查 GPU 设备名称。它应显示为 Intel Arc GPU。如果是，则说明您拥有符合我们最低规格要求的 GPU。如果显示为“Intel(R) Graphics”，则您的系统没有内置 Intel Arc GPU，不符合最低规格要求。如果您的 GPU 是独立显卡，例如 Intel Arc A 或 B 系列 GPU，则可以通过在设备管理器中禁用集成显卡来解决安装问题。\n4. **安装中断**：后端组件的在线安装可能会因 IT 网络、防火墙或休眠设置而中断或被阻止。请确保您处于开放网络环境中，关闭防火墙，并将休眠设置调整为接通电源时保持唤醒状态。\n5. **缺少库文件**：部分 Windows 系统可能缺少必要的库文件。可通过从 Microsoft [此处](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fcpp\u002Fwindows\u002Flatest-supported-vc-redist?view=msvc-170)安装 64 位 VC++ 再发行版来修复。建议在更新显卡驱动程序后再进行此操作，然后安装 AI Playground。\n6. **Python 冲突**：部分已安装 Python 的电脑可能会导致 AI Playground 安装冲突，原因是系统中 Python 的版本或位置不正确，从而安装了错误或冲突的包。通常可通过卸载 Python 环境、重启并重新安装 AI Playground 来解决。\n7. **临时文件**：如果由于上述任何问题导致安装中断，可能会留下临时安装文件，继续尝试安装时这些文件会阻止安装进程。请删除这些文件，或执行一次全新的 AI Playground 安装以解决问题。\n\n## 项目开发\n### 检出源代码\n\n要开始使用，请克隆仓库并进入项目目录：\n\n```cmd\ngit clone -b dev https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground.git\ncd AI-Playground\n```\n\n### 安装 Node.js 依赖项\n\n1. 从 [Node.js](https:\u002F\u002Fnodejs.org\u002Fen\u002Fdownload) 安装 Node.js 开发环境。\n\n2. 进入 `WebUI` 目录并安装所有 Node.js 依赖项：\n\n```cmd\ncd WebUI\nnpm install\n```\n\n### 获取外部资源\n\n1. 在 `WebUI` 目录下，执行 `fetch-external-resources` 脚本以下载所需的外部资源：\n\n这将下载 `uv`（Python 包管理器）及其他所需工具到 `build\u002Fresources\u002F` 目录。\n\n### 启动应用程序\n\n要在开发模式下启动应用程序，请运行：\n\n```\nnpm run dev\n```\n\n### （可选）构建安装程序\n\n要构建安装程序，请运行：\n\n```\nnpm run build\n```\n\n安装程序可执行文件将位于 `build\u002Felectron` 文件夹中。\n\n## 模型支持\nAI Playground 不随附任何生成式 AI 模型，但通过界面直接或用户从 HuggingFace.co 或 CivitAI.com 下载模型并将其放置在相应的模型文件夹中，可以为所有功能提供模型支持。\n\n当前应用中链接的模型\n\n### AI 模型与许可证注册表\n\n| 模型路径 \u002F 名称 | 模型卡片 (HF) | 许可证链接 |\n| :--- | :--- | :--- |\n| AdamCodd\u002Fvit-base-nsfw-detector | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FAdamCodd\u002Fvit-base-nsfw-detector) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Aitrepreneur\u002Finsightface\u002Finswapper_128.onnx | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FAitrepreneur\u002Finsightface) | [非商业用途](https:\u002F\u002Fhuggingface.co\u002FAitrepreneur\u002Finsightface#license) |\n| alimama-creative\u002FFLUX.1-Turbo-Alpha | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Falimama-creative\u002FFLUX.1-Turbo-Alpha) | [FLUX.1-dev 许可证](https:\u002F\u002Fhuggingface.co\u002Fblack-forest-labs\u002FFLUX.1-dev\u002Fblob\u002Fmain\u002FLICENSE.md) |\n| BGE Small EN v1.5 (GGUF) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FBAAI\u002Fbge-small-en-v1.5) | [MIT 许可证](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT) |\n| black-forest-labs\u002FFLUX.2-klein-4b-fp8\u002Fflux-2-klein-4b-fp8.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fblack-forest-labs\u002FFLUX.2-klein-4b-fp8) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| city96\u002Ft5-v1_1-xxl-encoder-gguf\u002Ft5-v1_1-xxl-encoder-Q3_K_M.gguf | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fcity96\u002Ft5-v1_1-xxl-encoder-gguf) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| city96\u002Ft5-v1_1-xxl-encoder-gguf\u002Ft5-v1_1-xxl-encoder-Q4_K_M.gguf | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fcity96\u002Ft5-v1_1-xxl-encoder-gguf) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| city96\u002Fumt5-xxl-encoder-gguf\u002Fumt5-xxl-encoder-Q4_K_M.gguf | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fcity96\u002Fumt5-xxl-encoder-gguf) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| comfyanonymous\u002Fflux_text_encoders\u002Fclip_l.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fcomfyanonymous\u002Fflux_text_encoders) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| comfyanonymous\u002Fflux_text_encoders\u002Ft5xxl_fp8_e4m3fn_scaled.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fcomfyanonymous\u002Fflux_text_encoders) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Comfy-Org\u002Fflux1-kontext-dev\u002Fflux1-dev-kontext_fp8_scaled.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002Fflux1-kontext-dev) | [FLUX.1-dev 许可证](https:\u002F\u002Fhuggingface.co\u002Fblack-forest-labs\u002FFLUX.1-dev\u002Fblob\u002Fmain\u002FLICENSE.md) |\n| Comfy-Org\u002FLumina_Image_2.0_Repackaged\u002Fae.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002FLumina_Image_2.0_Repackaged) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Comfy-Org\u002FReal-ESRGAN_repackaged\u002FRealESRGAN_x4plus.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002FReal-ESRGAN_repackaged) | [BSD-3-Clause](https:\u002F\u002Fopensource.org\u002Flicenses\u002FBSD-3-Clause) |\n| Comfy-Org\u002FWan_2.1_ComfyUI_repackaged\u002Fwan_2.1_vae.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002FWan_2.1_ComfyUI_repackaged) | [Wan 2.1 许可证](https:\u002F\u002Fhuggingface.co\u002FWan-AI\u002FWan2.1-T2V-14B\u002Fblob\u002Fmain\u002FLICENSE) |\n| Comfy-Org\u002Fz_image_turbo\u002Fae.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002Fz_image_turbo) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Comfy-Org\u002Fz_image_turbo\u002Fqwen_3_4b.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002Fz_image_turbo) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Comfy-Org\u002Fz_image_turbo\u002Fz_image_turbo_bf16.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002Fz_image_turbo) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| DeepSeek-R1-Distill-Qwen 1.5B | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fdeepseek-ai\u002FDeepSeek-R1-Distill-Qwen-1.5B) | [MIT 许可证](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT) |\n| DeepSeek-R1-Distill-Qwen 7B | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fdeepseek-ai\u002FDeepSeek-R1-Distill-Qwen-7B) | [MIT 许可证](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT) |\n| Gemma 3 4B IT (Unsloth) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Funsloth\u002Fgemma-3-4b-it) | [Gemma 许可证](https:\u002F\u002Fai.google.dev\u002Fgemma\u002Fterms) |\n| gmk123\u002FGFPGAN\u002FGFPGANv1.4.pth | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fgmk123\u002FGFPGAN) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| GPT-OSS 20B (Unsloth) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Funsloth\u002Fgpt-oss-20b) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| InternVL2 4B (OV) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FOpenGVLab\u002FInternVL2-4B) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| latent-consistency\u002Flcm-lora-sdv1-5\u002Fpytorch_lora_weights.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Flatent-consistency\u002Flcm-lora-sdv1-5) | [OpenRAIL++](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| latent-consistency\u002Flcm-lora-sdxl\u002Fpytorch_lora_weights.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Flatent-consistency\u002Flcm-lora-sdxl) | [OpenRAIL++](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Lightricks\u002FLTX-Video\u002Fltxv-2b-0.9.6-distilled-04-25.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FLightricks\u002FLTX-Video) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Llama 3.2 3B Instruct | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FLlama-3.2-3B-Instruct) | [Llama 3.2 许可证](https:\u002F\u002Fgithub.com\u002Fmeta-llama\u002Fllama-models\u002Fblob\u002Fmain\u002Fmodels\u002Fllama3_2\u002FLICENSE) |\n| lllyasviel\u002Ffooocus_inpaint\u002Ffooocus_inpaint_head.pth | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Flllyasviel\u002Ffooocus_inpaint) | [OpenRAIL](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| lllyasviel\u002Ffooocus_inpaint\u002Finpaint_v26.fooocus.patch | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Flllyasviel\u002Ffooocus_inpaint) | [OpenRAIL](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Lykon\u002FDreamShaper\u002FDreamShaper_8_pruned.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FLykon\u002FDreamShaper) | [OpenRAIL-M](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Lykon\u002Fdreamshaper-8-inpainting\u002Ftext_encoder\u002Fmodel.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FLykon\u002Fdreamshaper-8-inpainting) | [OpenRAIL-M](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Lykon\u002Fdreamshaper-8-inpainting\u002Funet\u002Fmodel.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FLykon\u002Fdreamshaper-8-inpainting) | [OpenRAIL-M](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Lykon\u002Fdreamshaper-8-inpainting\u002Fvae\u002Fmodel.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FLykon\u002Fdreamshaper-8-inpainting) | [OpenRAIL-M](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| Meta-Llama 3.1 8B Instruct | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fmeta-llama\u002FMeta-Llama-3.1-8B-Instruct) | [Llama 3.1 许可证](https:\u002F\u002Fgithub.com\u002Fmeta-llama\u002Fllama-models\u002Fblob\u002Fmain\u002Fmodels\u002Fllama3_1\u002FLICENSE) |\n| Mistral 7B Instruct v0.2 (OV) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FMistral-7B-Instruct-v0.2) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Mistral 7B Instruct v0.3 | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FMistral-7B-Instruct-v0.3) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Mistral 7B Instruct v0.3 (OV) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fmistralai\u002FMistral-7B-Instruct-v0.3) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Nomic Embed Text v1.5 (GGUF) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fnomic-ai\u002Fnomic-embed-text-v1.5) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Phi-3 Mini 4k Instruct (OV) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fmicrosoft\u002FPhi-3-mini-4k-instruct) | [MIT 许可证](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT) |\n| Phi-3.5 Mini Instruct (OV) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fmicrosoft\u002FPhi-3.5-mini-instruct) | [MIT 许可证](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT) |\n| QuantStack\u002FWan2.1_14B_VACE-GGUF\u002FWan2.1_14B_VACE-Q8_0.gguf | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FQuantStack\u002FWan2.1_14B_VACE-GGUF) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Qwen2-VL 7B Instruct (OV) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen2-VL-7B-Instruct) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Qwen3 4B (OV) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FQwen\u002FQwen2.5-4B) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Qwen3 4B (Unsloth) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Funsloth\u002FQwen2.5-4B) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Qwen3 4B Instruct 2507 (Unsloth) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Funsloth\u002FQwen2.5-4B-Instruct) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Qwen3-VL 4B Instruct (Unsloth) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Funsloth\u002FQwen2-VL-7B-Instruct) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| RunDiffusion\u002FJuggernaut-XL-v9\u002FRunDiffusionPhoto_v2.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FRunDiffusion\u002FJuggernaut-XL-v9) | [OpenRAIL-M](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FCompVis\u002Fstable-diffusion-license) |\n| SmolLM2 1.7B Instruct | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FHuggingFaceTB\u002Fsmollm2-1.7b-instruct) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| stabilityai\u002Fcontrol-lora\u002Frank128-canny-rank128.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fstabilityai\u002Fcontrol-lora) | [SAI Community 许可证](https:\u002F\u002Fhuggingface.co\u002Fstabilityai\u002Fcontrol-lora#license) |\n| tencent\u002FHunyuan3D-2.1\u002Fhunyuan3d-dit-v2-1\u002Fmodel.fp16.ckpt | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuan3D-2.1) | [Hunyuan3D 许可证](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuan3D-2.1\u002Fblob\u002Fmain\u002FLICENSE.txt) |\n| tencent\u002FHunyuan3D-2\u002Fhunyuan3d-dit-v2-0\u002Fmodel.fp16.safetensors | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuan3D-2) | [Hunyuan3D 许可证](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHunyuan3D-2\u002Fblob\u002Fmain\u002FLICENSE.txt) |\n| TinyLlama 1.1B Chat (OV) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002FTinyLlama\u002FTinyLlama-1.1B-Chat-v1.0) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n| Whisper (OV) | [模型卡片](https:\u002F\u002Fhuggingface.co\u002Fopenai\u002Fwhisper-large-v3) | [Apache 2.0](https:\u002F\u002Fwww.apache.org\u002Flicenses\u002FLICENSE-2.0) |\n\n请务必查看在 AI Playground 中使用的任何模型的许可条款，尤其要注意其中的限制性规定。\n\n\n\n### 使用替代模型\n有关详细信息，请参阅[用户指南](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground\u002Fblob\u002Fmain\u002FAI%20Playground%20Users%20Guide.pdf)，或观看此视频[链接](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=1FXrk9Xcx2g)，了解如何将替代的 Stable Diffusion 模型添加到 AI Playground 中。\n\n\n## 通知与免责声明：\n有关 AI Playground 的条款、许可及免责声明的信息，请访问 GitHub 仓库中的项目和文件：\u003C\u002Fbr >\n[许可证](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground\u002Fblob\u002Fmain\u002FLICENSE) | [通知与免责声明](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground\u002Fblob\u002Fmain\u002Fnotices-disclaimers.md)\n\n该软件可能包含第三方组件，这些组件附有单独的法律声明或受其他协议约束，具体说明可在随软件提供的“第三方通知”文件中找到。\n\n## 致谢\n所借用代码和组件的许可详情可在我们的 [3rdpartynoticeslicense](3rdpartynoticeslicenses.txt) 文件中找到。此外，以下机构及其工作对 AI Playground 的实现至关重要：\n* PyTorch - https:\u002F\u002Fpytorch.org\u002F \n* Stable Diffusion - https:\u002F\u002Fgithub.com\u002FStability-AI\u002Fstablediffusion\n* ComfyUI -  https:\u002F\u002Fgithub.com\u002Fcomfyanonymous\u002FComfyUI\n* OpenVINO - https:\u002F\u002Fopenvinotoolkit.github.io\u002Fopenvino.genai\u002F \n* Llama.cpp - https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp \n* Vue.js - https:\u002F\u002Fvuejs.org\u002F \n* 以及无数其他开源项目和贡献者，正是他们让这一切成为可能！","# AI-Playground 快速上手指南\n\nAI-Playground 是一款专为 Intel AI PC 设计的开源生成式 AI 应用套件。它支持聊天、代码辅助、文档搜索、图像分析及音视频生成等功能，所有计算均在本地离线运行，无需将数据上传至云端。\n\n## 环境准备\n\n### 系统要求\n*   **操作系统**: Windows\n*   **处理器 (CPU)**: Intel Core Ultra Series 3, Series 2H, Series 2V, 或 Series 1 H\n*   **显卡 (GPU)**: \n    *   内置 Intel Arc GPU (集成显卡)\n    *   或 独立 Intel Arc GPU (A 系列或 B 系列)，需具备 **8GB+ 显存 (vRAM)**\n    *   *注意：若设备管理器中仅显示 \"Intel(R) Graphics\" 而无 \"Intel Arc\" 字样，则不满足最低配置。*\n\n### 前置依赖 (开发者模式)\n若需从源码构建或开发，请确保安装以下环境：\n*   **Node.js**: 最新 LTS 版本 ([下载链接](https:\u002F\u002Fnodejs.org\u002F))\n*   **Git**: 用于克隆仓库\n*   **网络环境**: 首次运行或构建时需访问 GitHub 和 HuggingFace 下载后端组件及模型（建议确保网络通畅或使用代理）。\n\n## 安装步骤\n\n### 方式一：使用预编译安装包 (推荐普通用户)\n\n1.  **下载安装程序**\n    访问官方 Release 页面下载最新 beta 版安装包：\n    [AI-Playground-installer.exe](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Fdownload\u002Fv3.0.3-beta_hf\u002FAI-Playground-installer.exe)\n\n2.  **执行安装**\n    双击运行 `.exe` 文件。此步骤仅安装 Electron 前端，速度极快。\n\n3.  **初始化后端**\n    首次启动应用时，会弹出 \"AI Playground Setup\" 窗口。\n    *   保持网络连接畅通。\n    *   程序将自动下载并配置所需的后端组件（如 Llama.cpp, PyTorch 等），此过程可能需要数分钟。\n    *   完成后即可开始使用。\n\n### 方式二：源码构建 (推荐开发者)\n\n1.  **克隆仓库**\n    ```cmd\n    git clone -b dev https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground.git\n    cd AI-Playground\n    ```\n\n2.  **安装 Node.js 依赖**\n    ```cmd\n    cd WebUI\n    npm install\n    ```\n\n3.  **获取外部资源**\n    执行脚本以下载 `uv` (Python 包管理器) 及其他必要工具至 `build\u002Fresources\u002F` 目录：\n    ```cmd\n    npm run fetch-external-resources\n    ```\n    *(注：国内用户若下载缓慢，可检查脚本逻辑手动替换为国内镜像源，或等待官方后续支持)*\n\n4.  **启动开发服务器**\n    ```cmd\n    npm run dev\n    ```\n\n5.  **(可选) 构建安装包**\n    ```cmd\n    npm run build\n    ```\n    生成的安装程序位于 `build\u002Felectron` 文件夹中。\n\n## 基本使用\n\nAI-Playground 默认不包含模型文件，需在应用内下载或手动放置模型。\n\n### 1. 模型准备\n*   **自动下载**: 在应用界面中选择所需功能（如聊天或绘图），系统通常会提示并从 HuggingFace 自动拉取推荐模型。\n*   **手动部署**: 从 [HuggingFace](https:\u002F\u002Fhuggingface.co) 或 [CivitAI](https:\u002F\u002Fcivitai.com) 下载模型文件，将其放入应用对应的模型文件夹中。\n    *   支持格式包括 GGUF (LLM), Safetensors (Diffusion) 等。\n    *   推荐模型示例：`Qwen3 VL` (视觉对话), `Flux.1` (图像生成), `DeepSeek R1 Distilled` (推理)。\n\n### 2. 核心功能体验\n\n*   **智能对话 (Chat & Reasoning)**\n    *   选择加载了 LLM (如 `Qwen3 VL` 或 `DeepSeek R1`) 的会话。\n    *   直接输入文本问题进行问答。\n    *   **视觉分析**: 拖入图片，使用支持 Vision 的模型（如 Qwen3 VL）进行图像内容描述或分析。\n    *   **文档检索 (RAG)**: 导入本地文档，基于文档内容进行提问。\n\n*   **图像生成 (Image Generation)**\n    *   切换至图像生成模块。\n    *   选择模型（如 `SDXL`, `Flux.1`, `Z-Image`）。\n    *   输入提示词 (Prompt)，点击生成。\n    *   支持从低分辨率草稿到高质量成图的多种模式。\n\n*   **图像编辑 (Image Editing)**\n    *   导入图片，使用内置工具进行无损放大 (Upscaling)、局部重绘 (Inpainting) 或外绘 (Outpainting)。\n    *   所有编辑过程均在本地 GPU 加速完成，保护隐私且无订阅费用。\n\n> **提示**: 首次运行特定模型时，可能需要短暂时间进行初始化加载。请确保电源连接正常以避免因休眠导致中断。","独立游戏开发者小林正在为一款赛博朋克风格的游戏制作宣传素材，需要快速生成高质量的概念图并基于本地剧情文档进行角色对话测试。\n\n### 没有 AI-Playground 时\n- **数据隐私风险高**：将未公开的游戏设定文档和手绘草图上传至云端 AI 服务，担心核心创意泄露给第三方。\n- **工作流割裂严重**：生成图片需用 Stable Diffusion WebUI，聊天推理需打开浏览器访问网页版，切换多个工具且配置环境复杂。\n- **离线无法工作**：在网络不稳定或无网环境下，所有生成式 AI 功能立即瘫痪，导致创作中断。\n- **硬件算力闲置**：虽然配备了搭载 Intel Arc 显卡的 AI PC，但因缺乏统一调度软件，本地强大的 NPU 和 GPU 算力无法被有效利用。\n- **成本负担重**：高频次的图像生成和大模型推理需要购买昂贵的云端算力订阅服务，增加了项目预算压力。\n\n### 使用 AI-Playground 后\n- **数据完全本地化**：所有剧情文档分析与草图处理均在本地运行，敏感数据无需离开电脑，彻底消除泄露隐患。\n- **一站式全能工作台**：在一个界面内即可调用 Flux.1 生成高清概念图，同时利用 Qwen3 VL 模型分析画面并进行剧情对话，无需切换应用。\n- **随时随地离线创作**：依托 Intel Core Ultra 处理器与 Arc 显卡，即使断开网络也能流畅运行大模型，保证创作连续性。\n- **硬件性能满血释放**：自动适配并调用本地 Intel Arc GPU 加速推理，将原本闲置的硬件转化为高效的生成引擎。\n- **零成本自由探索**：摆脱云端订阅费用限制，可无限次尝试不同的图像风格（如 SDXL、Z-Image）和模型参数，降低试错成本。\n\nAI-Playground 让开发者在保护隐私的前提下，将本地 AI PC 转化为免费、离线且功能完备的私有化生成式 AI 工作站。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fintel_AI-Playground_d561605a.png","intel","Intel® Corporation","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fintel_bfc55d04.png","",null,"webadmin@linux.intel.com","https:\u002F\u002Fgithub.com\u002Fintel",[84,88,92,96,99,103,106,110],{"name":85,"color":86,"percentage":87},"TypeScript","#3178c6",52.4,{"name":89,"color":90,"percentage":91},"Vue","#41b883",37.8,{"name":93,"color":94,"percentage":95},"Python","#3572A5",5.5,{"name":97,"color":98,"percentage":23},"CSS","#663399",{"name":100,"color":101,"percentage":102},"JavaScript","#f1e05a",1.2,{"name":104,"color":80,"percentage":105},"NSIS",0.9,{"name":107,"color":108,"percentage":109},"HTML","#e34c26",0.1,{"name":111,"color":112,"percentage":113},"Batchfile","#C1F12E",0,809,88,"2026-04-03T02:34:14","MIT","Windows","必需：Intel Core Ultra 系列处理器（内置 Intel Arc GPU）或独立 Intel Arc A\u002FB 系列显卡，显存需 8GB 以上。不支持 NVIDIA CUDA。","未说明",{"notes":122,"python":123,"dependencies":124},"1. 仅支持 Windows 操作系统。2. 必须使用 Intel 架构硬件（Core Ultra 或 Arc 独显），首次运行需联网下载后端组件，耗时较长。3. 若系统已安装 Python 可能导致冲突，建议卸载后重试。4. 可能需要安装 Microsoft VC++ 64-bit 运行库。5. 遇到驱动问题可能需要使用 DDU 清理缓存。6. 模型文件不包含在安装包中，需通过界面下载或手动从 HuggingFace\u002FCivitAI 获取。","未说明（开发环境需安装 Node.js，运行时自动管理 Python 依赖）",[125,126,127,128,129],"PyTorch 2.10","Llama.cpp (Vulkan)","OpenVINO","Electron","Node.js",[26,14,52,54],"2026-03-27T02:49:30.150509","2026-04-06T06:44:30.870433",[134,139,144,149,154,159],{"id":135,"question_zh":136,"answer_zh":137,"source_url":138},12561,"为什么生成的图像（如 Z-image, Flux, QWEN）速度突然变得非常慢？","可以尝试以下解决方案：\n1. 升级到 3.0.2 beta 版本，该版本恢复了能改善体验的参数。\n2. 手动编辑 settings.json 文件（建议使用 VS Code 或 Notepad++，避免使用 Windows 自带记事本导致语法错误），添加显存保留参数。配置示例如下：\n{\n  \"debug\": false,\n  \"comfyUiParameters\": [\"--lowvram\",\"--reserve-vram\",\"6.0\"],\n  \"availableThemes\": [\"dark\", \"lnl\", \"bmg\", \"light\"],\n  \"currentTheme\": \"light\",\n  \"isDemoModeEnabled\": false\n}\n这有助于在显存不足时同时利用系统内存，提升稳定性。","https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Fissues\u002F389",{"id":140,"question_zh":141,"answer_zh":142,"source_url":143},12562,"AI Playground 启动时卡在加载界面无法进入怎么办？","这通常是因为缺少必要的运行库或安装顺序错误。请严格按照以下步骤操作：\n1. 卸载所有已安装的 AI Playground 相关文件。\n2. 重启电脑。\n3. 在安装软件前，先安装最新版的 Microsoft Visual C++ Redistributable (建议 64 位版本)。\n4. 再次重启电脑。\n5. 重新安装 AI Playground。\n6. 安装完成后直接启动，随后会提示下载模型文件，等待下载完成即可正常使用。此外，确保已更新至 v1.22 或更高版本，该问题已在后续版本中修复。","https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Fissues\u002F81",{"id":145,"question_zh":146,"answer_zh":147,"source_url":148},12563,"如何在 AI Playground 中使用自定义转换的 OpenVINO 模型？为什么转换后的模型不显示在列表中？","AI Playground 内置的 IPEX-LLM 会在模型加载时自动将其量化为 int4 精度（先加载到 CPU 量化，再加载到 GPU），因此通常不需要手动转换模型。如果手动转换的 OpenVINO 模型无法显示，可能是因为格式或量化方式不兼容。官方文档建议对于 OpenVINO 模型，使用 INT4_ASYM 或 INT4_SYM 格式能在不同设备上获得最佳的性能与准确度平衡。建议直接使用支持的原始模型格式，让软件自动处理量化，或者参考社区提供的已优化 OpenVINO 模型仓库。","https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Fissues\u002F213",{"id":150,"question_zh":151,"answer_zh":152,"source_url":153},12564,"安装时提示\"requirements-unknown.txt does not exist\"错误如何解决？","该错误通常是由于显卡驱动程序过旧或不匹配导致的。即使笔记本制造商（如 Medion）官网显示没有可用驱动，也建议强制安装最新的 Intel 显卡驱动程序。请使用\"Intel Driver & Support Assistant (Intel DSA)\"工具检测并安装适用于您 Intel Arc 显卡的最新官方驱动，安装完成后重启电脑再尝试安装 AI Playground。","https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Fissues\u002F175",{"id":155,"question_zh":156,"answer_zh":157,"source_url":158},12565,"为什么 AI Playground 不允许在管理员账户下运行？能否移除管理员权限检查？","出于安全考虑，开发团队决定不再允许 AI Playground 在\"所有用户 (All User)\"或管理员权限场景下运行。虽然用户可以创建标准账户来运行应用，但这是为了降低潜在的安全风险。目前官方暂无计划移除此限制，建议用户按照指引创建一个标准的非管理员用户账户专门用于运行该软件。","https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Fissues\u002F245",{"id":160,"question_zh":161,"answer_zh":162,"source_url":148},12566,"IPEX-LLM 是如何处理模型量化的？会影响性能吗？","IPEX-LLM 会在模型首次加载时自动进行量化处理。具体流程是：先将 safetensors 格式的模型加载到 CPU，自动量化为 int4 精度，然后再将量化后的版本加载到 GPU。这个过程每次首次运行模型时都会发生。这种自动量化机制不仅能显著减小模型体积，还能在保持性能的同时提高推理速度，因此用户无需预先准备量化版本的模型。",[164,169,174,179,184,189,194,199,204,209,214,219,224,229,234,239,244,249,254,259],{"id":165,"version":166,"summary_zh":167,"released_at":168},62936,"v2.3.0-alpha","This release is an alpha preview of AI Playground with support for Intel Core Ultra 200H Processors (ARL-H). This release is mostly functionally the same as version v2.2.1b, but replaces the Intel Extensions for PyTorch (IPEX) with native PyTorch 2.6, for compatibility with Intel Core Ultra 200H processors.  As this is our first release without IPEX, we are releasing this as an alpha version.\r\n\r\n> [!IMPORTANT]  \r\n> This alpha preview will run on all AI Playground supported skus. If performance or compatibility issues arise, notify us in our issues section and consider returning to version 2.2.1 (for supported systems).\r\n\r\nVersion 2.3a Features\r\n- All features from [AI Playground 2.2.1b](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Fv2.2.1-beta) \r\n- Support for Intel Core Ultra 200H (ARL-H) Processors\r\n- Image diffusion backend switch from IPEX 2.3 to PyTorch 2.6, eliminating IPEX libraries as a dependency\r\n- Generate Image Number field now editable to override the value limit beyond 4 images\r\n\r\nVersion 2.3a Fixes\r\n- UI\u002FUX adjustments including removing image history dimming during generation.\r\n- Non admin permissions enforced on AI Playground.exe\r\n\r\nKnown Issues\r\n- Version 2.3a with the PyTorch 2.6 backend can result in performance issues compared to 2.2.1b\r\n- Colorizer workflow is limited to B Series GPUs  \r\n- Installer can fail due environment issues.  See [Troubleshooting](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground#troubleshooting-installation) \r\n- PC restart may be needed after installation in order for optimal functionality and performance.\r\n- Custom changes to ComfyUI can break ComfyUI for AI Playground\r\n- Reinstalling over existing AI Playground will result in ComfyUI models and custom nodes being deleted\r\n- Users can adjust settings beyond their specific system's ability to generate or complete the task. \r\n\r\nVersion 2.3a Supported Hardware :\r\n- Intel Core Ultra 200H Processors (ARL-H)\r\n- Intel Core Ultra 200V Processors (LNL)\r\n- Intel Core Ultra 100H Processors (MTL-H)\r\n- Intel Arc B Series GPU Cards (BMG)\r\n- Intel Arc A Series GPU Cards with 8GB+ Memory (ACM)\r\n","2025-03-28T19:23:07",{"id":170,"version":171,"summary_zh":172,"released_at":173},62937,"v2.2.1-beta","This release includes needed fixes for v2.2, which introduced OpenVINO as an early preview as an high-performance backend for chat, v2.2 also includes additional ComfyUI worfkflows such for video and colorization.  This release provides a single installer for all supported hardware.\r\n_Intel Core Ultra 200-H (ARL-H) not yet supported._\r\n\r\n> [!IMPORTANT]  \r\n> Fixes In This v2.2.1 Release - Fixes a video generation issue from a failed installation of a needed component. Fixes image generation failing due to installation outside the C drive. Colorize works, but is limited to Xe2 processors (ie B580, B570, Intel Core Ultra 200v)\r\n\r\nBE SURE TO BACK UP VALUED CONTENT FROM A PREVIOUS VERSION BEFORE INSTALLING\r\n\r\nVersion 2.2 Features\r\n- All features from [AI Playground 2.0.4b](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Fedit\u002Fv2.0.4-beta) \r\n- **OpenVINO** chat backend early preview, using OpenVINO Gen AI, providing the most optimized inferencing solution for Intel hardware.  OpenVINO when installed can be used by selecting the Gear icon, then selecting it from the LLM Backend list in the Basic tab\r\n_NOTE this early preview does not yet have RAG implementation. This feature will be added in an upcoming release_\r\n- **Text To Video** Workflow.  Powered by LTX Video this workflow allows you to create video clips from text prompts \r\n- **Colorize Workflow**.  This workflow feature allows you to colorize black and white images\r\n_Limited to B Series GPUs and Intel Core Ultra 200v. Works best on people and faces_\r\n- Max token slider added, allowing you to set Answer response tokens up to 4096 tokens\r\n- UX improvements for Create tab providing full session history in the viewer with video player added for videos content\r\n- UX improvements for Answer section adding model information and performance summaries in the responses\r\n\r\nVersion 2.2 Fixes\r\n- Image output directory moved to User Directory folder, avoiding deletion of image output when updating or uninstalling AI Playground\r\n- Fixed ComfyUI workflow installation errors, incomplete installs, and erroneous online dependencies\r\n- Re-enable markdown files as RAG input\r\n- Installation timeout increased to reduce installation failures due to time-out\r\n\r\nVersion 2.2 Known Issues\r\n- Colorizer workflow is limited to B Series GPUs  \r\n- Installer can fail due environment issues.  \r\n_See [Troubleshooting](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground#troubleshooting-installation) section of the project Readme for installation troubleshooting_\r\n- PC restart may be needed after installation of AI Playground, otherwise can result in poor performance or functionality issues\r\n- Custom changes to ComfyUI can break the installation or repair of ComfyUI when installing AI Playground over an existing version.  \r\n_Users with custom ComfyUI set ups should backup and remove the ComfUI folder before installing a new version of AI Playground_\r\n- Users can adjust settings for image and video beyond their systems ability to produce or complete the task.  \r\n_Check for performance monitor to see if any usage is nearing your total GPU memory. If so you should decrease resolution and or frames for better performance._\r\n\r\nVersion 2.2 Supported Hardware :\r\n- Intel Core Ultra 200v Laptop Processors (LNL)\r\n- Intel Core Ultra 100h Laptop Processors (MTL-H)\r\n- Intel Arc B Series GPU Cards (BMG)\r\n- Intel Arc A Series GPU Cards with 8GB+ Memory (ACM)\r\n\r\nComing Soon\r\n- Intel Core Ultra 200h Laptop Processors (ARL-H)","2025-03-07T00:47:03",{"id":175,"version":176,"summary_zh":177,"released_at":178},62938,"v2.2-beta","This release introduces OpenVINO as an early preview and optional high-performance backend for chat, along with bug fixes, and additional ComfyUI worfkflows.  This release provides a single installer for all supported hardware.\r\n_Intel Core Ultra 200-H (ARL-H) not yet supported._\r\n\r\n> [!IMPORTANT]  \r\n> Two bugs sighted - affecting Video and Colorize Generation - fix in process, new release to come. [See issue](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Fissues\u002F211)\r\n\r\nBE SURE TO BACK UP VALUED CONTENT FROM A PREVIOUS VERSION BEFORE INSTALLING\r\n\r\nIn this release\r\n- All features from [AI Playground 2.0.4b](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Fedit\u002Fv2.0.4-beta) \r\n- OpenVINO backend early preview, using OpenVINO Gen AI, providing the most optimized inferencing solution for Intel hardware.  OpenVINO when installed can be used by selecting the Gear icon, then selecting it from the LLM Backend list in the Basic tab\r\n_NOTE this early preview does not yet have RAG implementation. This feature will be added in an upcoming release_\r\n- Text To Video Workflow.  Powered by LTX Video this workflow allows you to create video clips from text prompts \r\n- Colorize Workflow.  This workflow feature allows you to colorize black and white images - (works best on people and faces)\r\n- Max token slider added, allowing you to set Answer response tokens up to 4096 tokens\r\n- UX improvements for Create tab providing full session history in the viewer with video player added for videos content\r\n- UX improvements for Answer section adding model information and performance summaries in the responses\r\n\r\nFixes\r\n- Image output directory moved to User Directory folder, avoiding deletion of image output when updating or uninstalling AI Playground\r\n- Fixed ComfyUI workflow installation errors, incomplete installs, and erroneous online dependencies\r\n- Re-enable markdown files as RAG input\r\n- Installation timeout increased to reduce installation failures due to time-out\r\n\r\nSighted Bugs Since Release\r\n- Colorize Fails on Alchemist GPUs: Intel A series GPUs (ACM) get an 4GB memory limit issue: Issue is hijack fixes are missing in our ComfyUI backend implementation - a new release with fix is soon to publish. \r\n- LTX Video cannot complete: \"Backend could not generate image\" error to a required node not installing on some systems. Fix being added to AI Playground. **Workaround posted** here: https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Fissues\u002F206#issuecomment-2693282167\r\n\r\nKnown Issues\r\n- Installer may fail due to time, requiring a restart to finish.  \r\n_See [Troubleshooting](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground#troubleshooting-installation) section of the project Readme for installation troubleshooting_\r\n- PC restart may be needed after installation of AI Playground, otherwise can result in poor performance or functionality issues\r\n- Custom changes to ComfyUI can break the installation or repair of ComfyUI when installing AI Playground over an existing version.  \r\n_Users with custom ComfyUI set ups should backup and remove the ComfUI folder before installing a new version of AI Playground_\r\n- Users can adjust settings for image and video beyond their systems ability to produce or complete the task.  \r\n_Check for performance monitor to see if any usage is nearing your total GPU memory. If so you should decrease resolution and or frames for better performance._\r\n\r\nSupported Hardware :\r\n- Intel Core Ultra 200v Laptop Processors (LNL)\r\n- Intel Core Ultra 100h Laptop Processors (MTL-H)\r\n- Intel Arc B Series GPU Cards (BMG)\r\n- Intel Arc A Series GPU Cards with 8GB+ Memory (ACM)\r\n\r\nComing Soon\r\n- Intel Core Ultra 200h Laptop Processors (ARL-H)\r\n\r\n\r\n","2025-02-28T22:56:15",{"id":180,"version":181,"summary_zh":182,"released_at":183},62925,"v3.0.3-beta_hf","### **AI Playground 版本 v3.0.3 测试版 - 紧急修复**\n\n这是 AI Playground 3.0.3-beta 的紧急修复版本，用于解决已知的 bug 和版本兼容性问题。为确保与最新 OVMS 的兼容性，请将您的 NPU 驱动程序更新至最新版本。\n\n---\n### **紧急修复更新**\n* 更新了 OVMS 版本，以修复 GPU 上的 RAG 嵌入问题。\n* 更新了 Llama.cpp，以解决 Qwen3-4B 在推理输出方面的问题。\n* 由于与最新后端更新不兼容，已从 OpenVINO 模型列表中移除 Mistral 7B-instruct-v0.2-int4 和 Phi-3-mini-4k-instruct-gq。\n\n---\n### **版本 v3.0.3 测试版更新**\n详情请参阅：https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Ftag\u002Fv3.0.3-beta\n\n---\n\n### **已知问题**\n\n* 安装程序可能因环境问题而失败。请参阅 [故障排除指南](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002F#troubleshooting-installation) 获取帮助。\n* 在 Intel Core Ultra Series 3 处理器上，SDXL（HD、ControlFace、Sketch2Photo）工作流有时会出现随机黑屏输出。重新运行提示词和种子值可能会解决问题。该问题将在 PyTorch 2.12 中得到修复。\n* 具有独立显卡和集成显卡的系统在安装时可能需要禁用集成显卡。\n* 拥有多个显卡的系统在进行推理时，可能需要禁用空闲显卡。\n* 下载中断可能导致模型无法正确下载。若遇此情况，请前往模型所在目录，删除临时文件后重试。\n* Colorizer 工作流仅支持 Xe2 架构的显卡（LNL 和 BMG）。\n* 为获得最佳功能和性能，安装完成后可能需要重启电脑。\n* 在现有 AI Playground 安装基础上重新安装会删除所有 ComfyUI 节点。\n* 用户若将设置调整到超出其硬件能力的范围，可能导致生成失败。\n\n---\n\n### **版本 3.0.3 测试版支持的硬件**\n* Intel Core Ultra Series 3 处理器（PTL）\n* Intel Core Ultra Series 2 (H) 处理器（ARL-H）\n* Intel Core Ultra Series 2 (V) 处理器（LNL）\n* Intel Core Ultra Series 1 (H) 处理器（MTL-H）\n* Intel Arc B 系列显卡（BMG）\n* Intel Arc A 系列显卡（8GB 及以上显存，ACM）","2026-03-27T19:10:25",{"id":185,"version":186,"summary_zh":187,"released_at":188},62926,"v3.0.3-beta","### **AI Playground 版本 v3.0.3 测试版**\n\n这是 AI Playground 3.0.2-beta 的增量更新，包含更新的模型、后端、功能修复、安全补丁以及新增的操作模式。\n\n---\n### **版本 v3.0.3 测试版更新内容**\n\n* 图像生成手动预设现支持第三方 SD 1.5 和 SDXL 检查点及 LoRA 模型\n* 在聊天提示和加载图像设置工具中新增了摄像头拍摄功能\n* 后端更新：OpenVINO: 2026.1.0，ComfyUI: 0.17.0，Llama.cpp: b8429\n* Llama.cpp 现已支持 Qwen 3.5 9B 模型\n* NPU 聊天功能启用 RAG 嵌入\n* NPU 设备及模型中新增聊天指标\n* 更新了 NPU 模型以提升准确性\n* OpenVINO 模型支持 GPT-OSS 20B\n* 新增面向自助终端和参会者免费应用演示的演示模式。可通过 settings.json 启用。\n* 扩展了预设描述，并在提示设置中添加了信息图标详情\n* 当运行需要特定后端的功能时，新增缺失后端提醒\n* 输入图像现已加入历史记录并保存至磁盘，避免了本地存储问题\n* 多项 CVE 漏洞修复及安全更新\n---\n\n### **已知问题**\n\n* 安装程序可能因环境问题而失败。请参阅 [故障排除指南](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002F#troubleshooting-installation) 获取帮助。\n* Intel Core Ultra Series 3 处理器在使用 SDXL（HD、ControlFace、Sketch2Photo）时可能出现随机黑图。重新运行提示和种子值可能会解决问题。该问题将在 PyTorch 2.12 中修复。\n* 配备独立显卡和集成显卡的系统在安装时可能需要禁用集成显卡。\n* 具有多个显卡的系统在推理时可能需要禁用空闲显卡。\n* 下载中断可能导致模型无法正常下载。解决方法是前往模型所在目录，删除临时文件后重试。\n* Colorizer 工作流仅限于支持 Xe2 架构的 GPU（LNL 和 BMG）。\n* 为获得最佳功能和性能，安装后可能需要重启电脑。\n* 在现有 AI Playground 安装基础上重新安装会删除所有 ComfyUI 节点。\n* 用户若将设置调整到超出其硬件能力的范围，可能导致生成失败。\n\n---\n\n### **版本 3.0.3 测试版支持的硬件**\n\n* Intel Core Ultra Series 3 处理器（PTL）\n* Intel Core Ultra Series 2 (H) 处理器（ARL-H）\n* Intel Core Ultra Series 2 (V) 处理器（LNL）\n* Intel Core Ultra Series 1 (H) 处理器（MTL-H）\n* Intel Arc B 系列显卡（BMG）\n* Intel Arc A 系列显卡（ACM），且显存不低于 8GB","2026-03-25T16:40:24",{"id":190,"version":191,"summary_zh":192,"released_at":193},62927,"v3.0.2-beta","### **AI Playground 版本 v3.0.2 测试版**\n\n这是 AI Playground 3.0.1-alpha 的测试版本，修复了应用程序中的漏洞并添加了所需功能。\n\n---\n### **版本 v3.0.2 测试版更新内容**\n\n* 智能体聊天现在可以从对话中调用图像编辑预设工具\n* 智能体聊天的图像结果会显示在各自模式的历史记录中\n* 修复了聊天中的模型功能，支持功能标签\n* 为图像生成、图像编辑和视频生成模式新增了持久化历史记录\n* 图像历史记录现在会保存用于图像生成或图像编辑的“原始”参考图片\n* 专业预设升级至 Flux.2 Klein 模型\n* 控制人脸预设现包含使用 Z-Image-Turbo 模型的“质量”变体\n* 图像编辑结果中新增“发布到图像编辑模式”图标工具\n* 在生成额外内容的图像编辑预设中添加了安全检查功能\n* 修复了 SafetyChecker 离线超时的 bug\n* 修复了 dGPU 上的 Wan 2.1 VACE bug——已将预留显存重新添加到 ComfyUI 参数中\n* 在应用设置中新增 HuggingFace 镜像 URL 选项\n* 在 ComfyUI 后端设置选项中新增 ComfyUI 启动参数\n* ComfyUI 安装现在会自动将前端更新至最新版本\n* 在应用设置中新增持久化开发者控制台模式\n* 在应用设置中新增保持模型加载选项\n\n---\n\n### **已知问题**\n\n* 安装程序可能因环境问题而失败。请参阅[故障排除指南](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002F#troubleshooting-installation)获取帮助。\n* 在配备最新驱动程序的 Intel Core Ultra Series 3 处理器上，SDXL（高清、ControlFace、Sketch2Photo）可能会随机输出黑色图像。重新运行提示词和种子值或许可以解决。\n* 在较早世代及内存较低的 Intel Core Ultra 系统上，Gemma-3-4B 视觉模型可能出现问题。\n* 配备独立显卡和集成显卡的系统在安装时可能需要禁用集成显卡。\n* 拥有多个 GPU 的系统在推理时可能需要禁用空闲 GPU。\n* 下载中断可能导致模型无法正确下载。若遇此情况，请前往模型所在目录，删除临时文件后重试。\n* Colorizer 工作流仅适用于支持 Xe2 的 GPU（LNL 和 BMG）。\n* 为确保最佳功能与性能，安装后可能需要重启电脑。\n* 在现有 AI Playground 安装基础上重新安装会删除所有 ComfyUI 节点。\n* 用户若将设置调整至超出其硬件能力的范围，可能导致生成失败。\n\n---\n\n### **版本 3.0.2 测试版支持的硬件**\n\n* Intel Core Ultra Series 3 处理器（PTL）\n* Intel Core Ultra Series 2 (H) 处理器（ARL-H）\n* Intel Core Ultra Series 2 (V) 处理器（LNL）\n* Intel Core Ultra Series 1 (H) 处理器（MTL-H）\n* Intel Arc B 系列显卡（BMG）\n* Intel Arc A 系列显卡（8GB 及以上显存，ACM）","2026-02-13T20:20:38",{"id":195,"version":196,"summary_zh":197,"released_at":198},62928,"v3.0.1-alpha","### **AI Playground 版本 v3.0.1 alpha**\n\nAI Playground 3.0 早期 Alpha 预览版的次要更新。\n\n> [!重要]\n> 安装前请备份模型。\n---\n### **版本 v3.0.1 alpha**\n\n* 修复了在 Agentic 中勾选“启用工具”时视觉模型无法正常工作的问题\n* 修复了 Pro 2 预设中 VAE 模型的问题\n* 增加了编辑图像加载图片的优化\n* 调整了 ComfyUI 的参数\n* 切换到正确的 PyTorch 2.10 版本\n* 更新了依赖项\n\n---\n\n### **已知问题**\n\n* 配备独立显卡和集成显卡的系统，在安装时可能需要禁用集成显卡。\n* 下载中断可能导致模型无法正确下载。解决方法是前往模型所在目录，删除临时文件后重新尝试。\n* WAN 2.1 VACE 推理预览在中间步骤可能会显示为损坏，但最终输出结果是正常的。\n* 图像设置中的设备选择器可能会默认选择错误的设备，重启 AI Playground 后需要手动调整。\n* 如果不进行全新安装而直接覆盖旧版本的 AI Playground，可能会导致性能问题、输入提示时卡顿以及各种错误。\n* Colorizer 工作流仅限于支持 Xe2 架构的 GPU（LNL 和 BMG）。\n* 安装程序可能因环境问题而失败。有关帮助，请参阅[故障排除指南](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Ftree\u002F3.0.0-alpha#troubleshooting-installation)。\n* 为获得最佳功能和性能，安装后可能需要重启电脑。\n* 在现有 AI Playground 安装基础上重新安装会删除所有 ComfyUI 模型和自定义节点。\n* 用户可以将设置调整到超出其硬件实际处理能力的范围，从而导致任务无法生成或完成。\n\n---\n\n### **版本 3.0.1a 支持的硬件**\n* 英特尔 Core Ultra 系列 3 处理器（PTL）\n* 英特尔 Core Ultra 系列 2 (H) 处理器（ARL-H）\n* 英特尔 Core Ultra 系列 2 (V) 处理器（LNL）\n* 英特尔 Core Ultra 系列 1 (H) 处理器（MTL-H）\n* 英特尔 Arc B 系列显卡（BMG）\n* 英特尔 Arc A 系列显卡，且显存不低于 8GB（ACM）","2026-01-24T01:28:37",{"id":200,"version":201,"summary_zh":202,"released_at":203},62929,"v3.0.0-alpha","### **AI Playground 版本 v3.0.0 alpha**\n\n这是 AI Playground 3.0 的早期 alpha 预览版本。该版本已知存在一些 bug，但由于包含了重大变更和新功能，因此提前发布。欢迎提供反馈。\n\n有关此 alpha 版本的详细信息，请参阅 [AI Playground 3.0a 用户指南](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Fblob\u002F3.0.0-alpha\u002FAI%20Playground%20Users%20Guide.pdf)。\n\n> [!重要提示]\n> 本版本支持从代理工具调用到视频生成等广泛的功能。尽管属于例外情况，但在内存和计算资源较低的硬件上，部分功能可能会出现故障。\n\n---\n\n### **版本 v3.0.0 alpha**\n\n* 包含了[上一版本](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Ftag\u002Fv2.6.2-beta)（v2.6.2b）中的所有功能，除非在“修复与弃用”章节中另有说明。\n* **新增模态**\n  * 语音模式：AI Playground 现已支持语音模式。您可以在应用设置中开启“语音转文本”功能。这会将一个 AI 语音模型加载到您的 Intel Core Ultra PC 的 NPU 上。启用后，提示框中会出现一个麦克风图标。只需点击该图标，即可用语音输入提示内容。语音模式会将您的语音转换为文本提示。\n  * 视觉模型支持：AI Playground 新增了视觉模型支持。借助视觉支持，您可以将图片粘贴或导入到聊天模式中，让 AI Playground 分析该图片，为您提供关于图片的见解、答案和相关信息。\n  * 代理式多模态工具：AI Playground 将 MCP 工具引入聊天模式，使您能够在一次聊天对话中，使用核心模型跨 AI Playground 的各项功能执行代理式任务。此功能允许您在一个对话窗口内同时使用文本聊天、视觉聊天和图像生成等功能。更多信息请参阅代理式预设。\n* **统一提示框**：AI Playground 现在为所有功能采用单一的统一提示框，不再为每种模态单独设置标签页。这一设计使 AI Playground 的使用体验更加贴近常见的在线 AI 工具。\n* **浅色主题**：您现在可以选择以浅色主题体验 AI Playground。\n* **提示模式、设置与预设**：主提示框现在可以通过提示模式、提示设置和提示预设来定向至特定的模型和任务。\n  * 提示模式允许您在主提示框中快速便捷地切换不同模态：选择“聊天”以进行基于视觉和文本的聊天回复；选择“图像生成”以根据文本生成图像；选择“图像编辑”以修改或增强图像；选择“视频生成”以根据文本或图像输入创建视频内容。\n  * 提示模式预设：通过“提示设置”按钮，您可以访问数十种预设。这些预设类似于此前 AI Playground 中的“工作流”设置，专为特定任务和用途而配置，但现已扩展至所有提示模式。借助预设，您可以快速访问针对聊天、图像生成、图像编辑或视频生成的特定用法，其中模型选择和设置均已为该任务预先设定。S","2025-12-19T20:48:06",{"id":205,"version":206,"summary_zh":207,"released_at":208},62930,"v2.6.2-beta","### **AI Playground 2.6.2 版本**\n\n这是 AI Playground 的一个次要版本，主要修复了若干 bug 并优化了代码。\n\n> [!重要提示]\n> 本版本支持更多 NPU 模型。请务必在使用前[更新您的 NPU 驱动程序](https:\u002F\u002Fwww.intel.com\u002Fcontent\u002Fwww\u002Fus\u002Fen\u002Fdownload\u002F794734\u002Fintel-npu-driver-windows.html)，并重启电脑。\n>\n> 此外，本版本还支持 WAN 2.1 VACE 高级视频生成模型，这些模型在独立显卡上运行效果最佳。虽然它们也可以在内置 Intel Arc 显卡的 Intel Core Ultra 处理器上运行，但为了获得最佳性能，强烈建议使用独立显卡。\n\n---\n\n### **2.6.2b 版本功能**\n\n* 包含[上一版本](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Ftag\u002Fv2.6.1-beta)（v2.6.1b）的所有功能，除非在“修复与弃用”部分另有说明。\n\n---\n\n### **2.6.2b 版本修复与弃用**\n\n* 修复了 OpenVINO 与最新 PyTorch 2.9 之间的冲突，该冲突曾导致 OpenVINO 中无法进行文档嵌入。\n* 在 AI Playground 重启后，恢复各后端上次使用的推理设备。\n* 避免在安装软件包时加载全局 PIP 配置文件。\n* 移除基础设置中的过时设备选择选项。\n* 从构建中移除多余的 DLL 文件。\n* 更新 npm 依赖项。\n* 改进了 DLL 路径解析，并修复了默认后端和 ComfyUI 后端的 c10_xpu.dll 依赖性错误。\n* 修复了 Ollama PoC。\n\n---\n\n### **已知问题**\n\n* 下载模型和节点后首次运行时，部分功能可能会失败，需要重启 AI Playground。如果模型下载后立即出现故障，请重启 AI Playground 并重试。\n* 如果下载过程中断，可能会导致模型无法正常下载。解决方法是前往模型所在目录，删除临时文件，然后重新尝试。\n* 在推理预览过程中，WAN 2.1 VACE 的画面可能会显示异常，但最终输出结果是正常的。\n* 图像设置中的设备选择器有时会默认选错设备，需在重启 AI Playground 后手动调整。\n* 如果未进行全新安装而直接覆盖旧版本的 AI Playground，可能会导致性能问题、输入提示时卡顿以及各种错误。\n* Colorizer 工作流仅适用于支持 Xe2 架构的 GPU（LNL 和 BMG）。\n* 安装程序可能因环境问题而失败。有关帮助，请参阅[故障排除指南](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground#troubleshooting-installation)。\n* 为确保最佳功能和性能，安装后可能需要重启电脑。\n* 如果在现有 AI Playground 安装基础上重新安装，所有 ComfyUI 模型和自定义节点将会被删除。\n* 用户可以将设置调整到超出其系统实际能力的范围，从而导致任务无法生成或完成。\n\n---\n\n### **2.6.2b 版本支持的硬件**\n\n* Intel Core Ultra 200H 系列处理器（ARL-H）\n* Intel Core Ultra 200V 系列处理器（LNL）\n* Intel Core Ultra 100H 系列处理器（MTL-H）\n* Intel Arc B 系列显卡（BMG）\n* 具有 8GB 或以上显存的 Intel Arc A 系列显卡（ACM）","2025-10-22T17:02:22",{"id":210,"version":211,"summary_zh":212,"released_at":213},62931,"v2.6.1-beta","### **AI Playground 2.6.1b 版本**\n\n此版本对 AI Playground 进行了小幅修复和改进。\n\n> [!重要]\n> 本版本使用了新的签名密钥，这会重置 Microsoft 对该应用的信誉算法指标。下载时可能会提示该应用不受信任。请选择“保留”并确认信任该应用以绕过警告。\n>\n> 此版本支持更多 NPU 模型。请务必先[更新您的 NPU 驱动程序](https:\u002F\u002Fwww.intel.com\u002Fcontent\u002Fwww\u002Fus\u002Fen\u002Fdownload\u002F794734\u002Fintel-npu-driver-windows.html)，并在使用前重启电脑。\n>\n> 此版本还支持 WAN 2.1 VACE 高级视频生成模型，这些模型在独立显卡上运行效果最佳。虽然它们也可以在搭载内置 Intel Arc 显卡的 Intel Core Ultra 处理器上运行，但为了获得最佳性能，强烈建议使用独立显卡。\n\n---\n\n### **2.6.1b 版本功能**\n\n* 包含[上一版本](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Ftag\u002Fv2.6.0-beta)（v2.6.0b）的所有功能，除非在“修复与弃用”部分另有说明。\n* 为 NPU 设备上的 OpenVINO 增加了上下文大小设置。\n* 将 OpenVINO 更新至 2025.3，以更好地管理 Token 和上下文。\n* 向 Llama.cpp 添加了 Nomic 嵌入模型。\n* 升级了 Llama.cpp 的版本。\n* 改进了对话管理器的用户体验（@KyleHagy）——允许用户重命名对话。\n* 启用了通过 CTRL+鼠标滚轮缩放界面的功能，并可通过 CTRL+0 重置缩放。\n* 可根据[问题 315](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Fissues\u002F315) 配置 Hugging Face 端点：\n   * 通过 settings.json：默认值为 `\"huggingfaceEndpoint\": [\"https:\u002F\u002Fhuggingface.co\"](https:\u002F\u002Fhuggingface.co\u002F)`\n* 改进了后端安装及启动错误的显示：\n   * 在错误详情中添加失败后端的 `pip freeze` 输出。\n   * 捕获后端启动过程中的错误信息，并将其添加到错误详情中。\n\n\u003Cimg width=\"1049\" height=\"1093\" alt=\"image\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fc1b3a3ed-d49c-4478-93f8-ac35493d1c44\" \u002F>\n\n---\n\n### **2.6.0b 版本修复与弃用**\n\n* 通过将 insightface 的 stringzilla 依赖项固定为 3.12.6，修复了 CopyFace 和 FaceSwap 问题。\n* 移除了 BasicSR 组件，因为它已不再必要。\n* 重新打包了系统级 DLL 文件，以满足某些系统上的安装要求。\n\n---\n\n### **已知问题**\n\n* 下载模型和节点后，首次运行时部分功能可能会失败，需要重启 AI Playground。如果模型下载后立即出现故障，请重启 AI Playground 并再次尝试。\n* 如果下载过程中断，可能会导致模型无法正常下载。解决方法是前往模型所在目录，删除临时文件，然后重新尝试。\n* 在步骤过程中，WAN 2.1 VACE 推理预览可能会显示为损坏，但最终输出将是正常的。\n* 图像设置中的设备选择器可能会默认选择错误的设备，因此在重启 AI Playground 后需要手动进行设置。\n* 如果未执行干净安装而直接覆盖旧版本的 AI Playground，可能会导致性能问题、输入提示时出现卡顿以及各种错误。","2025-09-23T18:01:08",{"id":215,"version":216,"summary_zh":217,"released_at":218},62932,"v2.6.0-beta","### **AI Playground 2.6.0b 版本**\n\n此版本为 AI Playground 的“创作”和“回答”部分更新了新模型，并刷新了后端及 PyTorch 版本，以支持最新模型，例如 GPT-OSS。\n\n> [!重要]\n> 本版本支持更多 NPU 模型。请务必在使用前[更新您的 NPU 驱动程序](https:\u002F\u002Fwww.intel.com\u002Fcontent\u002Fwww\u002Fus\u002Fen\u002Fdownload\u002F794734\u002Fintel-npu-driver-windows.html)，并重启电脑。\n>\n> 此版本还支持 WAN 2.1 VACE 高级视频生成模型，这些模型在独立显卡上运行效果最佳。虽然它们也可以在内置 Intel Arc 显卡的 Intel Core Ultra 处理器上运行，但为了获得最佳性能，强烈建议使用独立显卡。\n\n---\n\n### **2.6.0b 版本功能**\n\n* 包含[上一版本](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Ftag\u002Fv2.5.5-beta)（v2.5.5b）的所有功能，除非在“修复与弃用”部分另有说明。\n* PyTorch 后端已更新至 PyTorch 2.8。\n* 新增 Flux.1 Kontext [dev] 的“按提示编辑”工作流。\n* 新增用于图像转视频和视频转视频的 WAN 2.1 VACE 工作流。\n* 添加了工作流下拉分类，使界面更整洁、访问更便捷。\n* 在图像设置中新增了“启动 ComfyUI”按钮。\n* 支持通过拖放将历史图片添加到工作流设置中。\n* 新增 Acer VisionArt 工作流，用于放大并把生成的图像放置到您的桌面上（需配备 Acer AI PC）。\n* 新增**演示模式**，为自助服务终端、演示和零售体验提供引导式提示。\n* Llama.cpp 后端已切换至 Vulkan 版本。\n* 在 Llama.cpp 中新增了 GPT-OSS 20B Q8 模型。\n* 使用 NPU 或 GPU 的 OpenVINO 增加了更多**压缩权重（cw）**模型。\n* 新增**后端版本控制**功能，允许用户调整所有已安装后端（ComfyUI、Llama.cpp 或 OpenVINO）的版本。\n* 在 Llama.cpp 中新增了**上下文大小**设置，用户可据此设置上下文，同时控制台会显示模型的内存使用情况。\n* 改进了 Markdown 格式。\n\n---\n\n### **2.6.0b 版本修复与弃用**\n\n* 已从 2.6.0 版本中移除 IPEX-LLM 支持。\n* “回答”部分的 Markdown 格式现已修复，聊天回答的格式得到改善。\n* Flux Schnell VAE 已切换为无限制版本，以减少下载错误。\n\n---\n\n### **已知问题**\n\n* 下载模型和节点后首次运行时，部分功能可能会失败，此时需要重启 AI Playground。如果模型在下载后立即出现故障，请重启 AI Playground 并重试。\n* 如果下载过程中断，可能会导致模型无法正常下载。要解决此问题，请前往模型所在位置，删除临时文件，然后再次尝试。\n* 在步骤执行过程中，WAN 2.1 VACE 推理预览可能会显示为损坏，但最终输出将是正常的。\n* 图像设置中的设备选择器可能会默认选择错误的设备，因此在重启 AI Playground 后需要手动进行设置。\n* 如果在未先清理旧版本的情况下直接覆盖安装 AI Playground，则…","2025-08-21T21:57:18",{"id":220,"version":221,"summary_zh":222,"released_at":223},62933,"v2.5.5-beta","这是一个小版本更新，对 AI Playground 2.5 进行了改进：新增了用于聊天的 NPU 设备选项，优化了图像生成的设备选择器，并修复了 2.5.0 版本中的一些问题。\n\n> [!重要]  \n> 本版本支持在聊天中使用 NPU。请务必先[更新您的 NPU 驱动程序](https:\u002F\u002Fwww.intel.com\u002Fcontent\u002Fwww\u002Fus\u002Fen\u002Fdownload\u002F794734\u002Fintel-npu-driver-windows.html)，然后重启电脑再使用。\n\n2.5.5b 版本特性：\n- 包含上一版本（v2.5.0b）的所有功能，包括对 Intel Core Ultra 200H (ARL-H) 处理器的支持。\n- 在“回答”选项卡中新增了设备选择器，并将图像生成的设备选择器移至“图像设置”选项卡。\n- “回答”选项卡中的 OpenVINO 后端现将 AI Boost（NPU）作为聊天选项，并支持文档嵌入功能。\n- IPEX-LLM 后端新增了一个嵌入模型：intfloat\u002Fmultilingual-e5-small。\n\n2.5.5b 版本修复内容：\n- 根据改进后的设备选择器，在配备独显的系统上安装时，现在可以启用集成显卡。\n- 视频工作流文件现已正确安装，不会出现云图标提示。\n- 主分支上的文本转视频文件已修正，移除了错误的图像输入字段。\n\n已知问题：\n- 安装模型和节点后，首次运行时部分功能可能会失败，可能需要重启 AI Playground。\n- “图像设置”中的设备选择器有时会默认选错设备，重启 AI Playground 后可能需要手动调整。\n- 如果不进行全新安装而直接在现有 AI Playground 上反复安装，可能会导致性能下降、提示框输入延迟以及各种错误。\n- Flux.1 Schnell VAE 模型现在需要 HuggingFace 用户访问令牌才能完成下载认证。请前往 HuggingFace 上该模型的页面登录，复制您的用户令牌，粘贴到 AI Playground 的“模型”选项卡中的用户令牌字段，以消除警告提示。\n- Colorizer 工作流仅适用于支持 Xe2 架构的 GPU（LNL 和 BMG）。\n- 安装程序可能因环境问题而失败，请参阅[故障排除指南](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground#troubleshooting-installation)。\n- 为获得最佳功能和性能，安装后可能需要重启电脑。\n- 在现有 AI Playground 上重新安装会删除 ComfyUI 中的模型和自定义节点。\n- 用户如果将设置调整得超出其硬件的实际处理能力，可能导致无法生成或完成任务。\n\n2.5.5b 版本支持的硬件：\n- Intel Core Ultra 200H 处理器（ARL-H）\n- Intel Core Ultra 200V 处理器（LNL）\n- Intel Core Ultra 100H 处理器（MTL-H）\n- Intel Arc B 系列显卡（BMG）\n- 具有 8GB 或以上显存的 Intel Arc A 系列显卡（ACM）","2025-05-23T17:15:59",{"id":225,"version":226,"summary_zh":227,"released_at":228},62934,"v2.5.0-beta","本次发布更新了 AI Playground，新增对 PyTorch 2.7 的支持，并包含一些小的修复和功能更新。\n\n> [!IMPORTANT]  \n> 本版本与最新版 Intel 显卡驱动程序配合使用时，在配备多块 GPU 的系统上可能会出现设备选择问题。如果您拥有独立的 Intel Arc GPU，且在安装或运行此版本的 AI Playground 时遇到问题，建议禁用您的集成 GPU（iGPU）。\n\n版本 2.5b 的新特性\n- 包含 [上一版本](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Ftag\u002Fv2.4.0-beta)（v2.4.0b）的所有功能，包括对 Intel Core Ultra 200H（ARL-H）处理器的支持。\n- 升级了对 PyTorch 2.7 的支持。\n- 升级了对 Python 3.12.10 的支持。\n- 将 OpenVINO 设置为聊天功能的默认后端。\n- 后端管理器和安装程序现在允许重新安装组件。\n- ComfyUI 安装程序现已默认安装 v0.3.30，并允许用户设置首选的 ComfyUI 版本。\n- 新增对 LTX-Video 0.9.5 和 0.9.6 模型及工作流的支持。\n\n版本 2.5b 的修复\n- 修复了 2.3a 和 2.4b 版本中存在的性能退化问题，提升了本版本的性能。\n\n已知问题\n- LTX-Video 工作流首次运行时可能会失败，需要在安装视频模型和节点后重启 ComfyUI 后端。\n- 较旧的 LTX-Video 工作流存在与 Python 3.12 不兼容的包问题，请使用标记为 v2.5 的新版 LTX-Video 工作流。\n- Flux.1 Schnell VAE 模型现在需要 HuggingFace 用户访问令牌进行身份验证才能下载。请访问 HuggingFace 上的模型页面，登录并复制您的用户令牌，然后将其粘贴到 AI Playground 的模型选项卡中的用户令牌字段，以消除警告提示。\n- 在某些环境中，设备设置可能会错误地被设置为 iGPU 而非 dGPU。作为临时解决方案，可以尝试禁用 iGPU。\n- Colorizer 工作流仅限于支持 Xe2 架构的 GPU（LNL 和 BMG）。\n- 安装程序可能因环境问题而失败。请参阅[故障排除](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground#troubleshooting-installation)部分。\n- 为确保最佳功能和性能，安装后可能需要重启电脑。\n- 如果在现有 AI Playground 安装基础上重新安装，ComfyUI 中的模型和自定义节点将会被删除。\n- 用户可能会将设置调整到超出其系统实际处理能力的范围，从而导致任务无法生成或完成。\n\n版本 2.5b 支持的硬件：\n- Intel Core Ultra 200H 处理器（ARL-H）\n- Intel Core Ultra 200V 处理器（LNL）\n- Intel Core Ultra 100H 处理器（MTL-H）\n- Intel Arc B 系列显卡（BMG）\n- 具有 8GB 或以上显存的 Intel Arc A 系列显卡（ACM）","2025-05-02T19:05:38",{"id":230,"version":231,"summary_zh":232,"released_at":233},62935,"v2.4.0-beta","This release updates AI Playground with fixes and features for chat, such as RAG embeddings for all backend options and DeepSeek models support for OpenVINO Gen AI.\r\n\r\nVersion 2.4.b Features\r\n- All features from [previous release](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Ftag\u002Fv2.3.0-alpha) (v2.3.0a) including support for Intel Core Ultra 200H (ARL-H) Processors\r\n- Document search and summary support for both OpenVINO and Llama.cpp back ends (RAG support)\r\n- DeepSeek R1 Distill model support for OpenVINO Gen AI back-end\r\n- Reasoning mode chat output is now collapsible or expandable.\r\n- Chat back-end selection, and other chat settings move to the horizontal tool-bar above the chat prompt, for easier access\r\n- Generate Image Number field now editable to override the value limit beyond 4 images (properly implemented in this version)\r\n\r\nVersion 2.4b Fixes\r\n- All user installation option removed, eliminating the administrative permission launch error\r\n\r\nKnown Issues\r\n- Version 2.3a and 2.4b with the PyTorch 2.6 backend can result in lower image generation performance compared to version 2.2.1b\r\n- Some systems may erroneously set the device setting to iGPU instead of dGPU. Disable iGPU as a workaround.\r\n- Colorizer workflow is limited to Xe2 supported GPUs (LNL, and BMG)  \r\n- Installer can fail due to environment issues.  See [Troubleshooting](https:\u002F\u002Fgithub.com\u002Fintel\u002Fai-playground#troubleshooting-installation) \r\n- PC restart may be needed after installation in order for optimal functionality and performance.\r\n- Custom changes to ComfyUI can break ComfyUI for AI Playground\r\n- Reinstalling over existing AI Playground will result in ComfyUI models and custom nodes being deleted\r\n- Users can adjust settings beyond their specific system's ability to generate or complete the task. \r\n\r\nVersion 2.4b Supported Hardware :\r\n- Intel Core Ultra 200H Processors (ARL-H)\r\n- Intel Core Ultra 200V Processors (LNL)\r\n- Intel Core Ultra 100H Processors (MTL-H)\r\n- Intel Arc B Series GPU Cards (BMG)\r\n- Intel Arc A Series GPU Cards with 8GB+ Memory (ACM)\r\n","2025-04-14T19:19:29",{"id":235,"version":236,"summary_zh":237,"released_at":238},62939,"v2.0.4-beta","This is an updated release of AI Playground 2.0 alpha with updates and fixes.  This release provides a single installer for all supported hardware including the latest Intel Arc B580 GPU.\r\n\r\nBE SURE TO BACK UP VALUED CONTENT FROM A PREVIOUS VERSION BEFORE INSTALLING\r\n\r\nIn this release\r\n- All features from [AI Playground 2.0a](https:\u002F\u002Fgithub.com\u002Fintel\u002FAI-Playground\u002Freleases\u002Fedit\u002Fv2.0.0a-preview) \r\n- CopyFace workflow added providing an image to image solution similar to FaceSwap\r\n- DeepSeek models added to drop down model list in the Answer tab for IPEX-LLM mode\r\n-- deepseek-ai\u002FDeepSeek-R1-Distill-Qwen-7B\r\n-- deepseek-ai\u002FDeepSeek-R1-Distill-Qwen-1.5B\r\n- Added High VRAM warning to workflows that usually are best run on systems with more than 8GB of VRAM\r\n- Updated language translations\r\n- By default Create is set to **Standard** resolution with **Fast** mode on, to provide the fastest initial experience for image generation\r\n- Higher resolution options added up to 0.5 megapixel for Default Mode, Standard Resolution ( ie SD 1.5)\r\n\r\nFixes\r\n- Missing language translations for parts of the UI have been added.\r\n- Negative prompt fixed can now be applied to Default mode where appropriate - does not work in Fast mode\r\n- Hyperlinks from Model Settings fixed\r\n- Safe Check toggle feature added back to Settings. This fixes an issue where image generation returns black images, being censored. Turning off this feature will bypass it \r\n- Broken FaceSwap and CopyFace installation fixed, due to a move the custom node repo being moved\r\n- VC++ redistributable check and installation added back to the installation\r\n- Llama.cpp default model set, matching the IPEX-LLM experience\r\n\r\nKnown Issues\r\n- When running and installing workflows like CopyFace for the first time, the installation process may take a long time and not fully execute. A restart of AI Playground may be needed to complete the installation.\r\n- PC restart may be needed after installation of AI Playground, otherwise can result in poor performance or functionality issues\r\n- Performance degradation can happen after running a few different high memory models. Restarting backends through Backend Installation Manager will restore performance\r\n- Network firewalls and protected networks can interrupt or block the installation process. Install using an open network.\r\n- Alternate Python and Git installations on the PC may conflict with installation.\r\n- VC++ may be detected and not installed by the installer but may be required to be installed if out of date for the hardware.","2025-02-04T21:08:01",{"id":240,"version":241,"summary_zh":242,"released_at":243},62940,"v2.0.0a-preview","This release is an early alpha preview of AI Playground 2.0.  This release provides a single installer for all supported hardware including the latest Intel Arc B580 GPU.\r\n\r\nIn this release\r\n- All features from AI Playground v1.22b\r\n- A single installation executable able to install on all currently AI Playground supported Intel Arc GPU hardware: Intel Core Ultra-H Series 1, Intel Core Ultra Series 2, Intel Arc Series A GPUs or Intel Arc Series B GPUs with 8GB+ of vRAM.\r\n- New runtime installer to manage the online installation and restart functions of required or optional backend components\r\n- New optional ComfyUI backend to support workflow mode for image generation\r\n- New optional Llama.cpp backend to support GGUF file support in the Answer chatbot section\r\n- Workflow flow modes leveraging curated ComfyUI workflows to be used in AI Playground\r\n- Workflow synch supporting download of latest workflows from the AI Playground project\r\n- Flux.1-Schnell workflows, leveraging either Q4 or Q8 quantization\r\n- Line2Image workflow using SDXL and ControlNet Canny models for fast generation of images influenced by the structure of another image\r\n- FaceSwap workflow using SDXL to control the likeness of an image based on a reference image (non commercial use only)\r\n- Add Model tool to easy add and install LLM models using hugging face IDs\r\n- Image resolution scale tool, to set the resolution in megapixels\r\n\r\nKnown Issues\r\n- PC restart needed before running, otherwise can result in poor performance or functionality issues\r\n- Performance degradation can happen after running a few different high memory models. Restarting backends through Backend Installation Manager will restore performance\r\n- Caching issues may cause image workflow settings to be entered in defaults.  If odd results are generated double check resolution and steps are correct for the model \r\n- Network firewalls and protected networks can interrupt or block the installation process. Install using an open network.\r\n- Alternate Python and Git installations on the PC may conflict with installation.","2024-12-20T01:05:17",{"id":245,"version":246,"summary_zh":247,"released_at":248},62941,"v1.22beta","**AI Playground Beta v1.22v Installer**.  This is the November 2024 security and bug fix release of AI Playground 1.2 beta for Windows. Currently this release supports Intel Core Ultra-H (MTL-H and Intel Core Ultra 200V (LNL).  Download the appropriate binaries below\r\n\r\n**Fixes for this release**: \r\n- Fix for the Hugging Face version error that causes the application to not get past the loading screen\r\n- Version updates for 3rd party components\r\n\r\n**Features for 1.2**: \r\n- Existing features of AI Playground 1.01b PLUS\r\n- New Theme UI - (For Intel Core Ultra 200V only)\r\n- New LLM selector prepopulated with 3 model options for download (Phi3-mini-4k-instruct, Qwen2-1.5B-Instruct, Mistral-7B-Instruct-v0.3)\r\n- New Conversation Manager to save and store chat session\r\n- Font Size Tool:  increase or decrease the font size in the Answer section.\r\n- New Aspect Ratio selector to more easily select the appropriate resolutions for specific aspect ratios\r\n- Multi-Language support now adding Korean with existing support for English and Chinese for the UI\r\n- Stable Diffusion 1.5 SafetyChecker warning for images flagged by safety checker\r\n- Resizable Window\r\n\r\n**Known Issues & Workaround**\r\n- First Initial launch can take more than a minute to install\r\n- Some existing installations of VC++ may not have needed DLLs.  If error on load screen we suggest installing the 64bit  VS C++ redist https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fcpp\u002Fwindows\u002Flatest-supported-vc-redist?view=msvc-170\r\n- Online installer can be interrupted by network firewalls or sleep mode. Be sure to install on an open network and set computer to not sleep during the installation\r\n- First run of AI Chatbot in Answer may be very long or hang. Restarting fixes and issue is generally not seen again\r\n- Generating HTML code from Answers may not render in the viewport, but do display in the console\r\n- Systems with low vRAM (8G) can experience slower than normal generation times in HD mode (SDXL)\r\n\r\n**NOTE**: \r\n_download the appropriate executable for your target hardware_","2024-11-07T22:41:45",{"id":250,"version":251,"summary_zh":252,"released_at":253},62942,"v1.21beta","**AI Playground Beta v1.21v Installer**.  This is the October 2024 release of AI Playground for Windows with support of the new Intel Core Ultra 200V series processors, code named Lunar Lake (LNL).\r\n\r\n**Features**: \r\n- Existing features of AI Playground 1.01b PLUS\r\n- New Theme UI - (For Intel Core Ultra 200V only)\r\n- New LLM selector prepopulated with 3 model options for download (Phi3-mini-4k-instruct, Qwen2-1.5B-Instruct, Mistral-7B-Instruct-v0.3)\r\n- New Conversation Manager to save and store chat session\r\n- Font Size Tool:  increase or decrease the font size in the Answer section.\r\n- New Aspect Ratio selector to more easily select the appropriate resolutions for specific aspect ratios\r\n- Multi-Language support now adding Korean with existing support for English and Chinese for the UI\r\n- Stable Diffusion 1.5 SafetyChecker warning for images flagged by safety checker\r\n- Resizable Window\r\n\r\n**Fixes and updates**: \r\n- Automated VC++ Redistributable installer for systems missing needed DLLs\r\n- Exception handing, providing error and log for issues starting up AI Playground\r\n- Port Selector - will automatically choose a free port, if default is used\r\n- Notification pop-up when selecting HD mode, alerting of generation times on systems with low VRAM\r\n- Smaller installer footprint - IPEX files now pulled from source\r\n\r\n**Known Issues & Workaround**\r\n- Some existing installations of VC++ may not have needed DLLs.  If error on load screen we suggest installing the 64bit  VS C++ redist https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fcpp\u002Fwindows\u002Flatest-supported-vc-redist?view=msvc-170\r\n- Python installation process can take more than 10 minutes or be hung by a poor internet connection. Resetting wifi or restarting install has been known to fix\r\n- System level Python installs, not part of a virtual environment, may interfere AI Playground Python and causing the app not to start. \r\n- First run of AI Chatbot in Answer may be very long or hang. Restarting fixes and issue is generally not seen again\r\n- Generating HTML code from Answers may not render in the viewport, but do display in the console\r\n\r\n**NOTE**: \r\n_download the appropriate executable for your target hardware_","2024-10-01T18:56:13",{"id":255,"version":256,"summary_zh":257,"released_at":258},62943,"v1.01beta.mtl","**AI Playground Beta Intel Core Ultra-H Installer**.  This is the initial launch version of AI Playground, packaged as a Windows installer for systems with an Intel® Core™ Ultra-H Processor with built-in Intel® Arc™ GPU.\r\n\r\n**Features**: \r\n- Run a variety of AI capabilities locally using your Intel Arc GPU\r\n- Text To Image Generation using either SD1.5 and SDXL Checkpoint, Models and LoRAs\r\n- Image to Image Generation with support for; Upscaling, Upscaling with Variation, Image to Image Generation with Variance, Inpainting, and Outpainting\r\n- Chatbot with RAG Embeddings to search, summarize, stylize generalized knowledge from an LLM model or from local text content.\r\n- Model swapping of PyTorch LLM, SD1.5, and SDXL models\r\n- **Multi-Language support, Chinese language added (via PR from user @Nuullll)**\r\n\r\n**Fixes and updates**: \r\n- Fixes online installer causing issues with LLM after 8\u002F05\u002F24\r\n- Updated some components version, for increased security\r\n- Fix for port conflicts on static port 9999\r\n- Code and UX fixes via community support requests\r\n\r\n**Known Issues & Workaround**\r\n- Some systems may hang on Loading Screen and can be resolved by installing VS C++ redist https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fcpp\u002Fwindows\u002Flatest-supported-vc-redist?view=msvc-170\r\n- Python installation process can take more than 10 minutes or be hung by a poor internet connection. Resetting wifi or restarting install has been known to fix\r\n- System level Python installs, not part of a virtual environment, may interfere AI Playground Python and causing the app not to start. \r\n- First run of AI Chatbot in Answer may be very long or hang. Restarting fixes and issue is generally not seen again\r\n- Generating HTML code from Answers may not render in the viewport, but do display in the console\r\n- SDXL image generation on many Intel Core Ultra systems may take exceedingly long due to the size of SDXL models which may be near or exceed the shared memory available to the built-in GPU\r\n- Issues sighted with when system has an additional dGPU - work-around\u002Ffix in process\r\n\r\n**NOTE**: \r\n_This version will not run on other Intel Arc GPUs, this will only run on the built-in GPUs for Intel Core Ultra-H systems_","2024-08-07T23:22:57",{"id":260,"version":261,"summary_zh":262,"released_at":263},62944,"v1.01beta","**AI Playground Beta Desktop Windows Installer**.  This is an updated beta version of AI Playground, packaged as a Windows installer for Intel Arc GPUs discrete with 8GB+ of VRAM.\r\n\r\n**Features**: \r\n- Run a variety of AI capabilities locally using your Intel Arc GPU\r\n- Text To Image Generation using either SD1.5 and SDXL Checkpoint, Models and LoRAs\r\n- Image to Image Generation with support for; Upscaling, Upscaling with Variation, Image to Image Generation with Variance, Inpainting, and Outpainting\r\n- Chatbot with RAG Embeddings to search, summarize, stylize generalized knowledge from an LLM model or from local text content.\r\n- Model swapping of PyTorch LLM, SD1.5, and SDXL models\r\n- **Multi-Language support, Chinese language added (via PR from user @Nuullll)**\r\n\r\n**Fixes and updates**: \r\n- Fixes online installer causing issues with LLM after 8\u002F05\u002F24\r\n- Updated some components version, for increased security\r\n- Fix for port conflicts on static port 9999\r\n- Code and UX fixes via community support requests\r\n\r\n**Known Issues & Workaround**\r\n- Some systems may hang on Loading Screen and can be resolved by installing VS C++ redist https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fcpp\u002Fwindows\u002Flatest-supported-vc-redist?view=msvc-170\r\n- Python installation process can take more than 10 minutes or be hung by a poor internet connection. Resetting wifi or restarting install has been known to fix\r\n- System level Python installs, not part of a virtual environment, may interfere AI Playground Python and causing the app not to start. \r\n- First run of AI Chatbot in Answer may hang. Restarting fixes and issue is generally not seen again\r\n- Generating HTML code from Answers may not render in the viewport, but do display in the console\r\n\r\n**NOTE**: \r\n_This version will not run on Intel Core Ultra. Intel Core Ultra version will be available as a separate installer_\r\n","2024-08-07T21:45:28"]