一个是默默关注Sora、OpenAI背后的"大东家"，另一个则是AI Agent里的"新秀"，两方在自家产品生态中上马文生视 频的时间点几乎是一样的。可见，微软这一步棋到底慢了多久。 从Sora代号第一次问世到现在，整体局面的发展路径很像："微软想要，OpenAI不给，但最后微软如愿以偿得到了 Sora的副产品"。 微软将Sora免费了，但却已经晚了。 前天，微软Bing宣布在其应用程序中推出 Bing 视频创作器（Bing Video Creator），该功能基于OpenAI的Sora模型，允 许用户通过文本提示词生成视频。这也是Sora首次面向用户免费开放。 就在昨天，Manus推出原生文生视频，嵌入进了自家Agent的工作流中。 为什么说它是Sora的副产品？因为微软上线的Bing 视频创作器从产品力和宣传上来讲，很难说得上是个完整的产品。 当微软终于宣布将它免费向用户开放时，这个消息并未掀起预期中的热潮，反而透着一股尴尬的迟到感。网友对这款 产品的"自来水"评价很差，甚至直言：我们已经有了可灵和Veo，为何还用Sora？ Sora，这个曾被OpenAI寄予厚望、甚至被誉为"AI视频领域的牛顿时刻"的模 ...

当代 AI 视频创作者有三件套：提示词、积分、以及抽卡。 继 Veo 3 刚刚掀起一轮小高潮后，Manus 也能生成视频了，功能挺全，经过实测，在 Agent 加持下， 支持图生视频、文生视频等标配功能。 该功能目前已经向 Basic、Plus 和 Pro 用户开放抢先体验。 先说结论，你要真指望它一句话秒出大片，那还是先降低心理预期。 高情商，不是不能用，只是抽卡的概率有些感人；低情商，用网友的话来说，花里胡哨，视频质量也有种 2011 年的美。 按照过往惯例，Manus 大概率也是套壳某家 AI 视频模型，但鉴于目前还没厂商认领，我们也不好断言，而经过一轮实测，我们也总结出几个特点： 图生视频：效果能打，但也随机抽卡 从体验上看，Manus 的图生视频明显要比文生视频靠谱得多。 我上传了一张威尔史密斯的照片作为参考，让其生成吃面的视频，效果还算可接受，风格统一、角色一致性尚可。 肤色和构图风格维持得比较好，相比于当前的视频主流模型，算得上是正常发挥。 并且，5 秒的视频仅扣了 44 积分，考虑到如果是普通用户，那么开通一个 Basic 账号，积分也足够用了。 抽卡严重，基本默认生成约 5 秒的「默剧」片段 ...

Core Viewpoint - Tencent has launched an open-source video generation tool called HunyuanVideo-Avatar, which allows users to animate characters from a static image and audio, creating lifelike interactions and performances [3]. Group 1: Technology Features - HunyuanVideo-Avatar acts as a "digital director," interpreting a static image and animating it based on the emotional tone of the audio [3]. - The tool eliminates the "internet celebrity face" issue by embedding the user's photo into the model, preserving original details like clothing folds and background lighting [4]. - It can extract emotional features from audio, allowing for nuanced facial expressions beyond simple lip-syncing [5]. - The technology enables multiple characters to interact independently, with natural eye contact and gestures, enhancing realism in performances [6]. Group 2: Application Scenarios - In e-commerce, the tool can create AI hosts for live streaming, using product images and promotional text to engage customers and drive sales [6]. - In music platforms, it allows for real-time performances by AI avatars, such as singing new songs or narrating stories in children's voices [7]. - For film production, directors can generate storyboard animations from simple sketches and voice scripts, streamlining the creative process [8]. Group 3: Technical Requirements - The minimum configuration for smooth operation is an NVIDIA RTX 3090 GPU with 24GB memory, while the recommended setup includes an NVIDIA A100 GPU with 80GB memory [9]. - Additional requirements include 64GB DDR4 RAM (minimum) and 500GB NVMe SSD storage [9].

2025.05. 26 本文字数：3653，阅读时长大约6分钟 导读 ： "瑕疵非常多，也很贵。" 作者 | 第一财经 刘晓洁 吕倩 "如果AI生成的角色拒绝相信他们是AI生成的，会怎么样？" 近日，海外博主用谷歌最新视频模型Veo 3生成的一些人物视频火了。在这些视频中，有一群人集体高 呼抗议"We're not prompts（我们不是提示词）"，还有一位男士举着手机自拍，背景是美妙的高山峡 谷，他指着身后，"你想说我背后的完美创造物，仅仅是0和1的结果，一串二进制代码，再无其他？这 不合理。" 当然台词和剧本是人创作的，但由AI生成的这些人物和场景都极具真实感，无论是光线在人脸上投下的 阴影与高光，还是人物的长相、口型，在阳光下眯起眼睛的神态都极为自然。配合Veo 3新的原生音频 生成功能，人们再一次惊呼"真实不存在了"。 事实是否真的如此，视频生成的GPT时刻终于来了吗？第一财经记者采访的Veo 3的使用者们并不这么 认为。AI Talk主理人、AIGC创作者汗青提到，Veo 3确实是很好的技术，但并没有网传那么夸张，例 如视频生成质量有提升但不惊艳，价格不低，现阶段对实际生产帮助还不大。 AIGC创 ...

Core Insights - Google has introduced its advanced video generation model, Veo 3, which can create videos with both visuals and dialogue generated entirely by AI [4][5] - The model allows users to describe characters, scenes, and specify dialogue and tone using natural language, marking a significant advancement in video generation technology [4][5] Group 1: Features of Veo 3 - Veo 3 can generate long videos seamlessly, showcasing its ability to maintain narrative flow and audio quality [13][14] - The model supports various creative applications, including generating rap lyrics and interactive cooking shows, demonstrating its versatility [2][6][7] - Users have already begun experimenting with the model, creating unique and humorous content, such as a dialogue between animated muffins [6][7] Group 2: Upgrades and Additional Features - Google has also upgraded Veo 2, introducing a "reference video" feature to maintain consistent video style and character appearance [15][16] - Additional functionalities include camera control, frame continuity, and the ability to add or remove objects within the video [18][19]

Group 1: Financial Performance - In 2024, the revenue from LED display control systems accounted for 46.17% of total revenue [3] - The gross profit margin for 2024 was 55.25%, an increase of 3% year-on-year [17] - The net profit margin remained stable despite a 40% increase in financial expenses due to exchange rate fluctuations [18] Group 2: Market Position and Product Development - The company plans to enhance its product offerings by focusing on Micro LED technology and custom solutions [3] - The video processing equipment revenue grew by 25% in 2024, but the gross margin decreased by 3% due to increased competition and raw material costs [32] - The company aims to maintain its market position by investing in advanced technologies and improving customer service [31] Group 3: Customer and Supply Chain Management - The accounts receivable turnover days increased by 5 days to 48 days, primarily due to extended payment terms from commercial display clients [16] - The company has a diversified supplier strategy to mitigate supply chain risks, particularly for chips and PCBs [18] - In 2024, the proportion of overseas revenue increased to 19.1%, with a focus on global market expansion [18] Group 4: Research and Development - R&D expenses increased by 18% in 2024, with a focus on AI video generation and edge computing technologies [29] - The company has a strong commitment to R&D, with a budget of 540 million yuan, significantly higher than industry peers [29] - The proportion of R&D personnel slightly decreased to 41.17% due to an increase in sales staff [27] Group 5: Environmental and Regulatory Compliance - The company’s environmental investment increased by 30% in 2024, reflecting its commitment to sustainability [20] - Government subsidies accounted for 12% of net profit, primarily from R&D grants and tax incentives [24] - The company actively participates in industry standard-setting to ensure product compliance and compatibility [25]

「Going Global 出海周报」 是创业邦推出的出海系列栏目，旨在为出海领域的创业者和投资人精选 出海大事件、海外大公司、投融资消息，本篇为栏目第 286 篇报道。 整理丨赵晓晓 本周（202 4 . 05 . 11 - 2025.05.17）出海大事件包括： TikTok被欧盟指控广告违规，最高可能面临年营业 额6%的罚款；Temu可能在美国恢复全托管模式；SHEIN在美国降低零售价；速卖通继续加码百亿补贴； 淘宝加速出海，哈萨克斯坦上线俄语版；阿里国际站加推美国专场大促；南洋国际物流集团停止运营； 美团 Keeta、蜜雪同一天宣布进入巴西市场；高盛预言：未来90天中国出口将爆火；美国对华小额包裹关 税据报低至30%等。 出海四小龙 TikTok 被欧盟指控广告违规，最高可能面临年营业额 6% 的罚款 5 月 15 日，欧盟指控 TikTok 违反《数字服务法》规定，没有提供有关广告内容、目标用户和广告 付费者的必要信息。该法案规定，互联网平台需要发布一个广告资源库，旨在让研究人员和用户检测 诈骗广告。 如果这一指控成立， TikTok 最高可能面临全球年收入 6% 的罚款。据 Oberlo 数据， ...

过去一年不断有人问我，"藏师傅有没有通过一个提示词生成整段视频的产品啊，我愿意付费"或者是"藏师 傅，我这里有口播稿和素材有没有能帮我剪辑的 AI 产品"。 我跟他们说的都是应该快了，马上就会有的，这次终于有了！ Medeo（ https://ai.medeo.app/create ）：创作者的专属AI视频工作室。 无论你有多少素材，哪怕只有一句话，他都能帮你生成一个带口播、音乐的完整视频。 这篇内容我会用几个案例来展示这个产品有多强大，另外会介绍一些使用技巧。 先来看一些案例 最基础的能力是你提供素材或者口播稿，他会帮你完成剪辑并生成视频。 非常适合资讯类或者对内容控制要求高的需求。 而且你可以要求他严格按照你提供的口播稿生成视频，也可以提供信息之后让他自己发挥。 比如下面这个左边就是我提供了 Dia CEO 的发言之后让他自己发挥的，右边就是让他精准根据口播稿生成的 视频。 我还提供了一些 Dia 的截图和视频，如果不够的话他还会自己寻找素材匹配进去，整个成本非常低。 当别的信息搬运者还在复制文字的时候，你直接一个链接丢进去，已经出视频了。 下面这个科普视频，我整个提示词就只有这一段话，没有任何干预，所有 ...

Core Viewpoint - The article discusses the rise of AI-generated video content, particularly focusing on a startup named Hedra, which has developed a technology for creating talking baby podcasts using AI-generated characters [1][2]. Group 1: Company Overview - Hedra was founded in 2023 and offers a web-based video generation and editing suite centered around its proprietary Character-3 model [1][5]. - The company completed a $32 million Series A funding round on May 15, led by Andreessen Horowitz, with existing investors participating [2][5]. - The CEO of Hedra, Michael Lingelbach, identified a market gap between companies like Synthesia and Runway, aiming to create longer dialogue scenes with greater control [2][5]. Group 2: Technology and Product Development - The Character-3 model, launched in March, has been a significant turning point for user growth and is expected to enable more customized AI character interactions [5][6]. - Hedra's technology allows users to integrate various models for video generation, including those for image and audio generation, enhancing the overall video production capabilities [7]. Group 3: Market Position and Competition - Hedra's competitors include Captions, Cheehoo, Synthesia, and HeyGen, with Hedra claiming its video characters are more expressive than those of its rivals [7]. - Andreessen Horowitz's Matt Bornstein noted that as AI-driven video generation evolves, more tools focusing on character, action, voice, and editing will emerge [7].

多模态 AI 技术正以前所未有的速度重塑内容创作领域。 从2024年 OpenAI Sora 点燃全球想象，到近期，吉卜力风图片席卷全网。这个一度被视为 AI 终极想象力边界 的领域，正以前所未有的速度冲破技术壁垒。 视频生成作为技术难度与应用潜力并存的关键环节，也吸引了全球范围内的广泛关注和投入。 在追求更长时长、更高分辨率、更惊艳视觉效果的同时，内容一致性难以保证、生成过程可控性不足、以及高 昂的计算成本等核心挑战，依然限制了其在专业领域、大众娱乐领域的规模化应用。 在此背景下，由生数科技研发的视频生成模型 Vidu，展现出一条差异化的发展路径。在多模态视频生成技术 的早期发展阶段，通过集中资源解决专业用户的核心痛点，如一致性、可控性、效率，建立起差异化优势和用 户基础，尤其是在动画等特定领域形成壁垒。 根据生数科技廖谦在近期访谈中的阐述，Vidu 的核心定位是"全球领先的AI内容生产平台 "，这也意味着 ，除 了追求基础生成能力的提升，也需要优先解决实际工作流中的关键痛点。 比如，生数科技敏锐的发现，纯粹的文生视频因为难以控制一致性，应用者并不多 。而 Vidu 推出的"参考 生"（Reference ...