![[eb8c96fa-3c9c-4282-8ca8-0aaad94c4ea6.mp3]]

腾讯混元开源新模型，简化数字人语音驱动视频生成

简报：

• 腾讯今日宣布开源混元语音数字人模型 HunyuanVideo-Avatar，该模型由腾讯混元视频大模型与腾讯音乐天琴实验室 MuseV 技术联合研发，仅需上传一张人物图像和一段音频，便可生成图中人物自然说话或唱歌的视频。
• HunyuanVideo-Avatar 支持头肩、半身与全身等多种景别，并具备多风格、多物种及双人场景能力，适用于短视频、电商、广告等创作领域。
• 模型可自动理解图片和音频内容，实现自然表情、唇形同步及全身动作，提升数字人视频生成的真实性和动态性，大幅降低内容制作难度与成本。
• 目前该模型单主体能力已开源，用户可在腾讯混元官网体验，支持上传不超过14秒的音频并生成相应视频，后续将继续拓展功能与开源范围。

相关链接：

• IT之家报道

• 体验入口

• 项目主页

Tencent Releases Open-Source Hunyuan Model, Simplifying Digital Human Voice-Driven Video Generation

Newsletter:

• Tencent announced today the open-source release of the HunyuanVideo-Avatar model, a digital human voice model co-developed by Tencent’s Hunyuan Video Large Model team and Tencent Music’s Tianqin Lab MuseV technology. Users can generate a video of a person naturally speaking or singing by simply uploading a single image of the character and a piece of audio. 🎤
• HunyuanVideo-Avatar supports various framing options, including head-and-shoulders, half-body, and full-body shots, along with multi-style, multi-species, and dual-person scene capabilities. It is suitable for creative fields such as short videos, e-commerce, and advertising. 📹
• The model automatically interprets image and audio content, delivering natural expressions, lip synchronization, and full-body movements. This significantly enhances the realism and dynamism of digital human video generation while greatly reducing the difficulty and cost of content creation.
• Currently, the single-subject functionality of the model has been open-sourced. Users can experience it on the Tencent Hunyuan official website, uploading audio clips of up to 14 seconds to generate corresponding videos. Further feature expansions and open-source scope are planned for the future. 🚀

Related Links:

• IT Home Report
• Experience Portal
• Project Homepage

OpenAI测试“以ChatGPT账号登录”功能，拓展至第三方平台

简报：

• OpenAI已在官网披露正在测试“使用ChatGPT账号登录第三方应用”的新功能，目前正征求开发者意见以接入自家应用。
• 此举旨在拓展ChatGPT用户基础，并为包括网购、社交平台、智能设备等更广泛消费者领域提供快速便捷的登录方式，与苹果、谷歌、微软等竞争。
• OpenAI已在开源AI编码工具Codex CLI中试行该登录方式，并为Plus与Pro用户赠送相应API额度鼓励体验。
• 目前该功能尚未公布正式上线时间，也未透露有多少企业已申请试点。

相关链接：

• IT之家报道

OpenAI Tests "Log in with ChatGPT Account" Feature, Expanding to Third-Party Platforms

Newsletter:

• OpenAI has announced on its official website that it is testing a new feature allowing users to "log in to third-party applications with a ChatGPT account." The company is currently seeking feedback from developers to integrate this into their own apps. 🖥️
• This move aims to expand the ChatGPT user base and provide a fast, convenient login method for a broader range of consumer sectors, including online shopping, social platforms, and smart devices, competing with giants like Apple, Google, and Microsoft. 🚀
• OpenAI has already piloted this login method in its open-source AI coding tool, Codex CLI, and is offering corresponding API credits to Plus and Pro users to encourage trial usage. 💻
• The official launch date for this feature has not yet been announced, nor has the number of companies applying for the pilot program been disclosed.

Related Links:

• IT Home Report

英伟达开放NVLink技术，强化AI生态布局并吸引全球资源

简报：

• 英伟达创办人黄仁勋在Computex演讲宣布，将原本封闭的NVLink芯片互联技术首次开放给半导体同业，助力更多厂商高效开发AI相关产品，进一步扩大英伟达AI生态圈。
• NVLink开放策略被认为有助于吸引原有竞争对手加入英伟达体系，同时进一步稳固其AI市场主导地位，并为公司带来新的成长动能。
• 黄仁勋积极在台布局AI基建，包括与台积电、鴻海等合作打造台首台AI超级电脑，并在台设立海外总部，加码投资与招募千名研发工程师，带来磁吸产业资源效应，也引发当地人才与资源竞争担忧。
• 随着英伟达不断扩展AI版图，台湾相关企业正加紧应对人才流动和资源排挤的挑战。

相关链接：

• NVLINK從封閉變開放…揭黃仁勳AI野心無極限
• 黃仁勳打造AI霸業！輝達磁吸台灣資源兩面刃

NVIDIA Opens NVLink Technology, Strengthens AI Ecosystem Strategy and Attracts Global Resources

Newsletter:

• NVIDIA founder Jensen Huang announced during his Computex speech that the previously proprietary NVLink chip interconnect technology will be opened to the semiconductor industry for the first time. This move aims to help more companies efficiently develop AI-related products, further expanding NVIDIA's AI ecosystem. 🌟
• The decision to open NVLink is seen as a strategy to attract former competitors into NVIDIA's ecosystem, while also reinforcing its dominant position in the AI market and driving new growth momentum for the company.
• Jensen Huang is actively building AI infrastructure in Taiwan, including collaborations with TSMC and Foxconn to create Taiwan's first AI supercomputer. Additionally, NVIDIA is establishing an overseas headquarters in Taiwan, increasing investments, and recruiting thousands of R&D engineers. This creates a magnetic effect on industrial resources but also raises concerns about local talent and resource competition. 💻
• As NVIDIA continues to expand its AI territory, related companies in Taiwan are urgently addressing challenges posed by talent mobility and resource allocation pressures. ⚙️

Related Links:

• NVLINK Shifts from Closed to Open… Revealing Jensen Huang’s Limitless AI Ambitions
• Jensen Huang Builds AI Dominance! NVIDIA’s Magnetic Pull on Taiwan Resources a Double-Edged Sword

Meta拆分AI部门以提升产品开发和AGI研究

简报：

• Meta公司宣布将其人工智能部门重组为两个新团队，分别为AI产品团队和AGI Foundations（基础研究）团队。
• AI产品团队侧重于开发消费级AI功能，涵盖Facebook、Instagram、WhatsApp等产品及新推出的独立AI应用，以加快产品落地和用户体验提升。
• AGI Foundations团队将着重推动长期AI基础技术和Llama大模型的优化，支持Meta在生成式AI领域的深度布局。
• 此次重组未涉及裁员，旨在提升Meta在激烈AI竞赛中的竞争力，与OpenAI、谷歌等公司抗衡。
• Meta近期还推出“Llama for Startups”计划，鼓励初创企业利用其AI技术，并在4月举办了LlamaCon活动展示其AI生态进展。

相关链接：

• Meta重组AI部门，聚焦消费级产品与AGI基础研究
• Meta重新调整AI团队结构 努力加速产品开发

Meta Splits AI Division to Boost Product Development and AGI Research

Newsletter:

• Meta has announced the restructuring of its artificial intelligence division into two new teams: the AI Products Team and the AGI Foundations Team.
• The AI Products Team will focus on developing consumer-grade AI features across products like Facebook, Instagram, WhatsApp, and newly launched standalone AI applications, aiming to accelerate product deployment and enhance user experience. 🚀
• The AGI Foundations Team will prioritize long-term AI foundational technologies and the optimization of the Llama large language model, supporting Meta’s deep investment in generative AI.
• This reorganization does not involve layoffs and is designed to strengthen Meta’s competitiveness in the intense AI race against companies like OpenAI and Google. 💪
• Recently, Meta also launched the “Llama for Startups” initiative to encourage startups to leverage its AI technology, and hosted LlamaCon in April to showcase progress in its AI ecosystem. 🌟

Related Links:

• Meta Reorganizes AI Division, Focusing on Consumer Products and AGI Research
• Meta Adjusts AI Team Structure to Accelerate Product Development

WordPress组建AI团队，推动社区人工智能开发一体化

简报：

• 2025年5月28日，WordPress宣布正式成立AI团队，由Automattic AI负责人James LePage领衔，成员包括谷歌的Felix Arntz、Pascal Birchler和10up的Jeff Paul。
• 新团队将引导和协调WordPress开发者社区的AI产品研发，提升AI项目系统化和平台战略水平，减少分散，确保发展方向契合WordPress长期目标。
• 截至目前，WordPress社区开发者已发布超过660款AI插件，涵盖网站模板生成、搜索引擎优化和聊天机器人等领域。

相关链接：

• WordPress成立AI团队，推进插件平台战略
• AIBase新闻报道

WordPress Forms AI Team to Drive Community AI Development Integration

Newsletter:

• On May 28, 2025, WordPress announced the official formation of its AI team, led by Automattic AI head James LePage, with members including Felix Arntz and Pascal Birchler from Google, as well as Jeff Paul from 10up. 🤝
• The new team will guide and coordinate AI product development within the WordPress developer community, enhancing the systematization and platform strategy of AI projects, reducing fragmentation, and ensuring alignment with WordPress's long-term goals. 🚀
• As of now, WordPress community developers have released over 660 AI plugins, covering areas such as website template generation, search engine optimization, and chatbots. 💻

Related Links:

• WordPress Establishes AI Team to Advance Plugin Platform Strategy
• AIBase News Report