时间有了，却错过节点。 从剧本、拍摄到后期，每个环节都繁琐冗长，光来回沟通就能耗掉半条命。 预算有了，却产不出量。 "千人千面"的广告理想很美好，但现实是多做一个版本就得多掏一笔钱。 营销人常常自嘲： 创意有了，却拍不出来。 危险镜头不敢拍，奇幻场景找不到，特效还贵得离谱。 结果，视频广告成了品牌营销中"最好用却最难用"的利器：效果拔群，但又贵又慢又难规模化。然而，一场技术革命正悄然改变这一切。 8月21日，百度蒸汽机（MuseSteamer）音视频一体化模型完成重大升级，在行业内首次实现多人有声音视频一体化生成。创作者输入脚本，几分钟后你就 能收获一条有角色、有对白、有情感、有镜头语言的视频成片。更重要的是，它不是实验室里的"炫技模型"，而是已经在一汽-大众、伊利倍畅等品牌的 真实营销案例中落地，帮他们把原本需要动辄数十万、上百万预算的视频，变成几乎"零成本"的创意成片。 在品牌营销的世界里，视频一直是最贵又最折腾的内容拼图。 拍一条像样的TVC广告，可能要烧掉上百万元；一个节日创意短片，从立项到上线往往拖上一个月甚至更久。可等视频终于"千呼万唤始出来"，下一波热 点早把消费者刷屏了，品牌只能尴尬地发现自己 ...

机器之心PRO · 会员通讯 Week 34 --- 本周为您解读 ② 个值得细品的 AI & Robotics 业内要事 --- 1. 视频生成 vs 空间表征，世界模型该走哪条路？ 视频预测生成的高质量画面，是否真的意味着模型理解了物理与因果规律？直接在潜在空间建模能否有效避免像素噪声干扰，同时保持决策与规划能力？混合路线是否能成为未来世界模型的 最优路径？随着生成模型和潜在表征技术的发展，AGI 的「思想实验沙盒」能否真正落地应用于物理世界任务？... 2. 抢天才还是拼算力？前 Llama 推理负责人详解 AI 的真实天花板 真正决定 AI 行业天花板的，是天才研究员的灵感，还是指数级增长的算力？如果算力增长放缓，AI 行业会否面临「增长乏力」的拐点？高阶概念想法，如果没有系统实验验证，能否真正推 动模型跃迁？模型泛化的天花板，到底靠升级模型，还是靠设计更高质量的新考题？... 本期完整版通讯含 2 项专题解读 + 30 项本周 AI & Robotics 赛道要事速递，其中技术方面 12 项，国内方面 8 项，国外方面 10 项。 本期通讯总计 20464 字，可免费试读至 9% 消耗 288 微信 ...

Meta is diving into AI video generation with a splash. The company will work with and license models from Midjourney AI, one of the most popular AI image and video companies.Alexandr Wang, Meta's chief AI officer, revealed the partnership in a post on X Friday. It's still unclear when a possible Meta x Midjourney model could be available for people to use.Meta teased a possible tool, MovieGen, at its 2024 Connect event, but we haven't heard much since. Right now you can upload an existing file or image, or ...

参与过电影《2012》、《黑客帝国3》等好莱坞大片的视效指导姚骐今天公布了他用AI制作的科幻短片《归途》。 短片里，如同末日的世界里，巨大的异形生物追击驾驶汽车的人类、巨型蜘蛛爬在高楼等场景栩栩如生。姚骐评价"（效果）跟实拍差不 多。" 他向第一财经等记者透露，整部短片用了40多个镜头，每个镜头生成3次，共计120个视频片段，其中包括18个10秒一体化的有声片段和102个 五秒片段，最终花费约一周时间制作完成。 姚骐说，如果这部短片是一部纯实拍或者CG制作的片子，可能需要几百万的成本。在好莱坞做镜头，有些复杂的镜头仅一个就要几十万甚至 上百万。此外，实拍还受限于场景实现难度、危险性以及演员、剧组成本，而AI技术的介入为创意实现提供了全新可能。 几百万实拍成本的短片，如果用AI生成，花费是多少？ 姚骐AI短片的合作对象、百度商业体系商业研发总经理刘林告诉记者，该片使用百度蒸汽机音视频一体模型，整体成本约在330.6元人民币。 但AI生成还有不少进步空间 目前，百度视频生成模型上线50天，最大的用户来自百度内部，包括搜索业务、移动生态创作者等，其次是专业领域创作者，以及企业客 户。 眼下视频生成赛道已足够卷。快手 ...

Core Viewpoint - The article discusses the significant advancements brought by Google's DeepMind with the release of Genie 3, which showcases a new path towards Artificial General Intelligence (AGI) through the concept of "World Models" [4][5][6]. Group 1: Introduction of Genie 3 - On August 5, Google DeepMind launched Genie 3, a model capable of generating interactive 3D virtual environments based on user prompts, demonstrating enhanced real-time interaction capabilities compared to previous AI models [5]. - Genie 3 features a "Promptable World Events" function, allowing users to dynamically alter the generated environment through text commands, showcasing its advanced interactivity [5]. Group 2: Concept of World Models - World Models are inspired by the human brain's ability to create and utilize an "inner world" to simulate future scenarios, which is crucial for decision-making and action [8][9]. - The development of World Models has evolved from early attempts to mimic human cognitive functions to more sophisticated models that can predict and simulate real-world dynamics [10][11]. Group 3: Technical Implementation of World Models - The implementation of World Models involves several key stages: Representation Learning, Dynamic Modelling, Control and Planning, and Result Output, each contributing to the AI's ability to understand and interact with the world [15][16][17][18]. - Representation Learning allows AI to compress external data into an internal language, while Dynamic Modelling enables the simulation of future scenarios based on actions taken [15][16]. Group 4: Applications of World Models - World Models can significantly enhance "embodied intelligence," allowing AI agents to learn through simulated experiences in a safe environment, reducing costs and risks associated with real-world trials [20][21]. - In the realm of digital twins, World Models can create proactive simulations that predict changes and optimize processes in real-time, enhancing automation and decision-making [21][22]. - The education and research sectors can benefit from World Models by creating virtual laboratories for precise predictions and interactive learning environments [22]. Group 5: Potential and Challenges of World Models - While World Models present vast potential for various applications, they also raise ethical and governance concerns, such as the blurring of lines between reality and virtuality, and the potential for behavioral manipulation [24][25][26]. - The debate surrounding World Models as a pathway to AGI highlights differing opinions within the AI community, with some experts advocating for their necessity while others question their effectiveness compared to model-free approaches [28][29][30].

Core Viewpoint - The rise of AI-generated eating broadcasts (AI Mukbang) is not just a novelty but represents a new wealth opportunity, showcasing strong addictive qualities and becoming a significant source of income for content creators [6][7][8]. Group 1: AI Mukbang Phenomenon - AI Mukbang videos have gained immense popularity, with some accounts rapidly amassing hundreds of thousands of followers within days, creating significant traffic miracles [7][11]. - The hashtag AI ASMR on platforms like Xiaohongshu has over 5,000 notes, with related topics exceeding 1 million views, indicating high engagement levels [11]. - Creators are leveraging AI tools to produce content that transcends traditional eating broadcasts, allowing for imaginative scenarios that captivate viewers [24][25]. Group 2: Monetization Strategies - Content creators are monetizing their AI Mukbang videos through platform incentives and by selling AI prompt templates, which have become a valuable commodity in the ecosystem [17][21]. - A TikTok creator has reportedly earned their first income by selling a four-sentence prompt for $9.9, demonstrating the lucrative nature of prompt selling [17]. - Companies like PromptBase are emerging, allowing users to purchase prompts for $1.99, with the platform taking a 20% commission on transactions [21]. Group 3: Market Dynamics - The current landscape shows that companies creating AI tools are not profiting as much as those selling the tools, indicating a shift in the monetization model within the AI content creation space [28]. - For instance, Kuaishou reported over 150 million yuan in revenue from its AI tool, Keling, highlighting the financial potential of AI-generated content [28]. - The user base for Keling has surpassed 45 million, with a significant portion of revenue coming from professional users who are willing to pay for advanced features [29].

文/陈永伟 8月5日，谷歌DeepMind发布了其新模型——Genie 3。 该模型能够根据用户的文本或图像提示，实时生成可供用户与AI智能体（AI Agent）互动的3D虚拟环 境。例如，用户只需输入"月球上的火山边"，Genie 3便能即时生成一片浮动的火山、黄色的大地与远 处的宇宙背景，并允许用户进入探索。 相比此前的AI模型，Genie 3展现出更强的实时交互能力，并在互动时长和记忆连贯性上表现尤为出 色。例如，如果用户在生成的房间墙壁上涂鸦，然后转身探索别处，那么当他稍后返回时，墙上的涂鸦 依旧保留。 不仅如此，Genie 3还引入了"可提示的世界事件"（Promptable World Events）功能。这允许用户在交 互过程中，通过新的文本指令动态改变世界。无论用户要求"加入一只奔跑的小狗""把天气从晴天变成 大雨"，还是"将环境从海边变成山上"，Ge-nie 3都能瞬间响应。 Genie 3的出色表现不仅刷新了AI生成世界的边界，也让人们看到了另一条通向通用人工智能（AGI）的 路径——"世界模型"（World Model）的希望。一时间，关于"世界模型"的讨论频频见诸媒体。 那么，什么是" ...

Core Viewpoint - The rise of AI-generated eating broadcasts (AI Mukbang) represents a new wealth opportunity, showcasing strong addictive qualities and generating significant traffic and income for creators [7][11][30]. Group 1: AI Mukbang Phenomenon - AI Mukbang videos have gained immense popularity, with creators rapidly increasing their follower counts and engagement metrics on platforms like TikTok and Xiaohongshu [11][15][16]. - The content of AI Mukbang is diverse, featuring imaginative food items that do not exist in reality, appealing to viewers' desires for relaxation and escapism [26][27]. - The trend has led to a new monetization model where creators sell AI prompt templates, allowing others to generate similar content [19][22]. Group 2: Monetization and Business Models - Creators are leveraging platform incentives and selling courses on how to create AI-generated content, establishing a lucrative revenue stream [19][24]. - Companies like PromptBase have emerged, facilitating the sale of AI prompts, with a commission structure that benefits both the platform and the sellers [22][31]. - The AI content creation model is proving to be more profitable than traditional AI development, as platforms like Kuaishou report significant revenue from AI tools [31][32]. Group 3: Market Dynamics - The AI Mukbang trend is beneficial for short video platforms, as it drives user engagement and monetization opportunities for both creators and the platforms themselves [31][32]. - The user base for AI tools is expanding, with a significant portion of revenue coming from professional content creators who are willing to pay for advanced features [32]. - The success of AI Mukbang indicates a shift in content consumption, where viewers seek low-energy, relaxing experiences, further solidifying the demand for such content [27][30].

Core Viewpoint - The article discusses the rapid advancements in AI technologies, particularly in the realm of interactive entertainment, highlighting the emergence of AI-native startups that redefine content, social interactions, and entertainment forms [2][3]. Group 1: AI in Interactive Entertainment - The integration of AI in gaming and interactive entertainment is becoming a central topic among players and investors, as seen at the ChinaJoy exhibition [4]. - Users have high expectations for new interactive entertainment forms, with traditional gaming being gradually deconstructed by AI, leading to faster content consumption and higher demands for emotional value [4][5]. - The blending of AI with gaming, video, and social elements is deepening, driven by advancements in AI-native technologies and large model capabilities [6]. Group 2: Startup Insights and Product Development - Startups like Feeling AI are focusing on creating products that facilitate user-generated 3D content, emphasizing co-creation between AI and users [8][10]. - The company aims to allow users to create unique characters and narratives, fostering social interactions and collaborative storytelling [9][10]. - The importance of understanding user needs and narrative demands is highlighted, with a focus on structured storytelling as a core product feature [11]. Group 3: Future Trends and Market Fit - The article emphasizes the need for startups to find their Product-Market Fit (PMF) by experimenting across various sectors, including gaming, e-commerce, and education [30][31]. - The evolution of AI technologies is expected to redefine industry standards, with a call for companies to embrace innovative models and maintain agility in their approaches [37]. - The future of interactive entertainment is envisioned as a space where users can engage in immersive experiences, potentially transforming traditional content consumption into collaborative creation [40][41].

Core Insights - The global AI startup funding has reached an astonishing $122 billion since the beginning of the year, with the US market accounting for $104.3 billion, representing 85.5% of the total raised [1] - The AI funding landscape continues to grow, with a projected $110 billion in 2024 and significant investments from major players like Meta and Anduril [1][5] - Despite a slight decrease in total investment from the previous quarter, AI-related funding remains at historically high levels [4][5] Investment Trends - In Q2, global AI startup funding totaled $50 billion, nearly half of the total VC investment of approximately $101.5 billion during the same period [1][5] - The largest funding round this quarter was Meta's $14.3 billion investment in Scale AI, which resulted in CEO Mark Zuckerberg acquiring a 49% stake [5] - There is a notable shift towards AI projects with "intensive infrastructure," supported by significant public and private sector investments [6] Market Dynamics - The AI-driven venture capital market has shown resilience, with a year-over-year growth of 7.28% from 2023 to 2024 and 9.26% from 2024 to 2025, totaling a 17.22% increase over two years [5] - Major VC firms like SoftBank, Andreessen Horowitz, and Sequoia continue to dominate the AI startup funding landscape [7] - The concentration of capital in leading AI startups has created a challenging environment for smaller companies seeking funding [7] Future Projections - OpenAI plans to invest trillions in core AI infrastructure, including AI chips and advanced power systems, indicating a long-term commitment to AI development [8] - Analysts predict that major tech companies will spend over $350 billion on AI infrastructure in 2023, with expectations of nearly 50% growth in 2024 [8] - Morgan Stanley forecasts that the AI investment boom could add $13 to $16 trillion in value to the S&P 500 index, representing a potential 30% increase [9][10]