11月19日晚，快手举行2025年第三季度业绩电话会。会上，针对市场高度关注的视频生成赛道竞争格局 及可灵AI的下一步迭代方向，快手科技创始人兼首席执行官程一笑进行了回应。 程一笑表示，当前视频生成赛道涌现出众多来自互联网大厂与创业公司等不同类型的参与者，这体现出 视频生成是一个极具潜力的优质赛道，也意味着行业仍处在快速技术迭代和产品形态探索的早期阶段。 更重要的是，整个行业正在通过竞争加速进步，推动视频生成技术更好地满足用户需求，渗透更多应用 场景。 11月19日，快手披露的2025年三季报显示，三季度可灵AI营业收入超3亿元，其全球用户规模突破4500 万，累计生成超2亿个视频和4亿张图片。 面对快速演进的竞争格局，可灵AI保持着持续的技术与产品创新。今年9月底，可灵推出2.5 Turbo模 型，在文本响应、动态效果、风格保持、美学质量等多个维度实现大幅提升。 程一笑表示，可灵的愿景是"让每个人都能用AI讲出好故事"，公司将聚焦于AI影视创作这一核心目标， 聚合资源深入打磨技术与产品能力。在具体迭代方向上，可灵将围绕技术领先性与产品想象力双线推 进，围绕多模态交互理念（如MVL），结合用户需求洞察与技术突 ...

Core Viewpoint - The article introduces VAFlow, a novel framework for video-to-audio generation that directly models the mapping from video to audio, overcoming limitations of traditional methods that rely on noise-based priors [6][9][29]. Background - The transition from "noise to sound" to "video to sound" highlights the evolution in multimodal generation tasks, particularly in video-to-audio (V2A) generation [3]. Traditional Methods - Early V2A methods utilized autoregressive and mask-prediction approaches, which faced challenges due to the discrete representation of audio leading to quality limitations [4][5]. VAFlow Framework - VAFlow eliminates the dependency on Gaussian noise priors, enabling direct generation of audio from video distributions, resulting in significant improvements in generation quality, semantic alignment, and synchronization accuracy [6][8][9]. Comparison of Generation Paradigms - The article contrasts traditional diffusion models and flow matching methods with VAFlow, demonstrating that VAFlow achieves better performance in terms of convergence speed and audio quality metrics [19][20]. Prior Analysis - The study compares Gaussian prior and video prior, showing that video prior offers better alignment with audio latent space, leading to superior generation quality [12][15]. Performance Metrics - VAFlow outperforms existing state-of-the-art (SOTA) methods in audio generation quality metrics, achieving the best scores in various benchmarks without complex video conditioning modules [24][25]. Visual Results - The article presents visual comparisons of generated audio from VAFlow against ground truth, illustrating its capability to accurately interpret complex scenes and maintain audio-visual synchronization [27]. Future Directions - The research team plans to explore VAFlow's applications in broader audio domains, including speech and music, indicating its potential for general multimodal generation [29].

Summary of the Conference Call Company and Industry Involved - **Company**: 富博集团 (Fubo Group) - **Industry**: Generative AI, Video Content Creation, Copyright Management Key Points and Arguments Sora 2 Overview - Sora 2 allows users to generate videos through text or image prompts, similar to TikTok, with initial content quality being well-received. However, copyright management has tightened, limiting daily video generation and restricting prompts involving well-known IPs [2][3][4] - Sora 2 represents a new milestone in generative AI, with significant improvements in generation effects, image control, and clarity, creating investment opportunities in content production and computing power [2][5] Market Impact - The launch of Sora 2 has led to a surge in downloads, surpassing GenAI and ChatGPT, indicating strong market interest [3] - The application has made notable improvements in copyright protection, implementing stricter measures as user numbers grow, which is expected to enhance its position in the global market [4][7] Technological Advancements - Sora 2's technology marks a shift towards multimodal generation, requiring higher computational power compared to traditional models, thus presenting new challenges and opportunities in the market [6][9] - The Diffusing Transformer used in video generation faces memory constraints, necessitating large HBM or future DDR5 support, highlighting the need for advanced hardware [9] Business Growth and Strategy - Fubo Group anticipates significant business growth from Sora's overseas expansion, particularly in copyright management, through partnerships with major platforms [7][8] - The Solo 2 product, launched as an independent app, achieved 140,000 downloads in its first two days, indicating strong user demand for AI-generated video content [12] Future Collaborations and Trends - Future collaborations for Solo 2 may include partnerships with IP owners and content creators, expanding into various video formats and social media platforms [13] - The rise of AI-generated content is expected to increase the demand for copyright protection and management, impacting multiple related industries [11][25] Financial Outlook - Fubo Group's revenue is projected to grow significantly, with AI-generated content expected to dominate active assets by the end of the year [33] - The company recently completed a financing round of 1.6 billion HKD to support R&D and team expansion, positioning itself for future growth [34] Challenges and Opportunities - The evolving copyright landscape poses challenges, but also opportunities for companies like Fubo Group to establish themselves as leaders in copyright management and content generation [19][20] - The potential for traditional media companies like Disney to adapt to digital content trends could reshape the industry, emphasizing the importance of flexible copyright strategies [35][38] Conclusion - Fubo Group is well-positioned to leverage advancements in generative AI and video content creation, with a focus on copyright management and strategic partnerships to drive future growth and innovation in the industry [44][45]

Core Insights - The article discusses the Genesis framework, a multimodal image-point cloud generation algorithm developed by Huazhong University of Science and Technology and Xiaomi Auto, which does not require occupancy (OCC) guidance for generating realistic driving scene data [2][4]. Group 1: Genesis Framework Overview - Genesis employs a two-stage architecture: the first stage uses a perspective projection layout and scene descriptions to learn 3D features, while the second stage converts multi-view video sequences into a bird's-eye view feature space [4]. - The framework introduces DataCrafter, a data annotation module based on Visual Language Models (VLM), to provide structured semantic information for guiding the generation process [10][13]. Group 2: Challenges in Current Driving Scene Generation - Existing methods primarily focus on single-modal data generation, either RGB video or LiDAR point clouds, which limits the potential for deep collaboration and consistent expression between visual and geometric modalities [7][8]. - The high cost of obtaining OCC labels in real-world driving scenarios restricts the application of existing multimodal generation models in the industry [8]. Group 3: DataCrafter Module - DataCrafter is designed to filter training data and extract structured semantic information, ensuring high-quality segments are used for training and providing detailed semantic guidance for the generation tasks [13][18]. - The module evaluates video segments based on visual attributes such as clarity, structural coherence, and aesthetic qualities, retaining only those that meet a set threshold [15]. Group 4: Video Generation Model - The video generation model within Genesis integrates scene layout information and language descriptions through attention mechanisms, enhancing the semantic expression of dynamic scenes [19]. - Innovations include using YOLOv8x-Pose for detecting pedestrian poses, which are then projected across various views to improve the generation of realistic driving scenarios [19]. Group 5: Performance Metrics - In experiments on the nuScenes dataset, Genesis achieved a multi-frame FVD of 83.10 and a multi-frame FID of 14.90 without initial frame conditions, outperforming previous methods [26]. - For LiDAR generation, Genesis demonstrated superior performance with a Chamfer distance of 0.611 at 1-second prediction, surpassing the previous best by 21% [27]. Group 6: Downstream Task Evaluation - The generated data from Genesis was evaluated in downstream perception tasks, showing improvements in mean Average Precision (mAP) and NuScenes Detection Score (NDS) across various settings [30]. - The combination of camera and LiDAR modalities in generation tasks yielded the highest gains, demonstrating the complementary advantages of multimodal generation [30].

Core Viewpoint - The future of AI is moving towards multimodal generation, enabling the creation of high-quality video content from simple text or image inputs, significantly reducing the time and resources required for creative work [2][4][30]. Group 1: AI Video Generation Technology - xAI's Grok 4 emphasizes video generation capabilities, showcasing a full-chain process from text or voice to image and then to video [2]. - Baidu's MuseSteamer 2.0 introduces a groundbreaking Chinese audio-video integration model, achieving millisecond-level synchronization of character lip movements, expressions, and actions [4][5][6]. - The new model allows users to generate high-quality audio-visual content with just a single image or text prompt, marking a significant leap in AI video generation technology [5][30]. Group 2: Product Features and Pricing - MuseSteamer 2.0 offers various versions (Turbo, Lite, Pro, and audio versions) tailored to different user needs, with competitive pricing at only 70% of domestic competitors [8][10]. - The Turbo version generates 720p resolution videos in 5 seconds for a promotional price of 1.4 yuan, enhancing cost-effectiveness for users [8][10]. Group 3: User Experience and Testing - Users can experience the model through various platforms, including Baidu Search and the "Huixiang" application [12][15]. - Initial tests demonstrate that the AI-generated dialogues and actions are fluid and realistic, with high-quality synchronization between audio and visual elements [19][22][30]. Group 4: Technical Advancements - The model addresses two core challenges: temporal alignment of audio and video, and the integration of multimodal features to ensure natural interactions [31][32]. - Baidu's model has been trained on extensive multimodal datasets, focusing on Chinese language capabilities, which enhances its applicability for local creators [36][37]. Group 5: Market Impact and Future Prospects - The MuseSteamer 2.0 model is designed to meet practical application needs, integrating deeply into Baidu's ecosystem to enhance creativity and productivity for users and businesses [41][44]. - The cost of producing high-quality video content has drastically decreased, allowing more creators to participate in professional-level video production [44][46].

Core Insights - Tencent officially launched the Hunyuan 3D World Model 1.0 at the World Artificial Intelligence Conference on July 27, 2025, marking the industry's first open-source immersive, interactive, and realistic world generation model [1][3] - The model significantly simplifies the 3D scene construction process for game developers, allowing for quick generation of complex scenes from simple text prompts or images [3][9] - Tencent's commitment to open-source development is evident as it plans to release a series of smaller models and frameworks, enhancing community engagement and collaboration [16][18] Group 1: Hunyuan 3D World Model 1.0 Features - The Hunyuan 3D World Model 1.0 integrates panoramic image synthesis and layered 3D reconstruction technology, enabling high-quality, diverse 3D scene generation from text or image inputs [1][9] - Users can create complete 3D scenes, including architecture, terrain, and vegetation, with simple commands, which can be used for game prototyping and level design [3][9] - The model's innovative algorithm allows for semantic hierarchical representation and generation of 3D scenes, facilitating intelligent separation of foreground and background elements [9][13] Group 2: Model Performance and Community Engagement - The Hunyuan 3D World Model 1.0 outperforms leading open-source models in aesthetic quality and instruction adherence, establishing a strong position in the global market [13][16] - Tencent's Hunyuan models, including TurboS and T1, are rapidly evolving, with monthly updates enhancing their capabilities in code generation, mathematical reasoning, and text writing [14][18] - The company has embraced open-source principles, with over 2.3 million downloads of its 3D models, making it one of the most popular open-source 3D model platforms globally [18]

Core Viewpoint - The article discusses the capabilities of Nano AI in generating complete videos from a single sentence, highlighting its high success rate and versatility in creating various types of content such as news introductions, educational videos, and narrative summaries [3][14]. Group 1: Video Generation Capabilities - Nano AI has introduced a feature that allows users to generate complete videos from a single sentence, demonstrating impressive success rates [3]. - The system can create videos based on prompts, including detailed visual effects and narrative hooks to engage viewers [3][12]. - The process involves analyzing existing videos to generate new creative ideas, enhancing the quality and effectiveness of the output [6][10]. Group 2: Technical Process - The video generation process includes several steps: generating image prompts, creating voiceovers, producing video content, adding subtitles, and integrating music [11]. - The AI checks the output for quality and can regenerate any problematic elements, ensuring a polished final product [11][12]. - Currently, the voice matching for multiple characters is limited, but the overall style and presentation of the videos are noted to be engaging and humorous [12]. Group 3: Future Potential - The article emphasizes that the trend for the year is towards code generation and multimodal generation, with complete video automation being a significant milestone [14]. - As the capabilities of large language models (LLMs) and video/audio models improve, the potential for video generation agents is expected to expand significantly [14]. - The current limitations in audio and voice processing are anticipated to be resolved with the introduction of new models, leading to a breakthrough in video generation technology [14].

Core Viewpoint - The article discusses the potential of multimodal generative AI, particularly in the advertising sector, highlighting its successful applications and the opportunities it presents for talent in this field [3][4][11]. Group 1: Current State of AIGC and Multimodal Generation - The job market for narrow AIGC roles, such as video generation, appears limited, leading to concerns about employment prospects for those with backgrounds in foundational vision and generative models [2][3]. - Despite the early stage of technology development, multimodal generation has already seen successful applications in advertising, yielding tangible benefits for major companies [3][4]. Group 2: Generative AI in Advertising - Generative AI has been utilized in advertising for years, with platforms like Amazon launching AI tools to enhance content generation, significantly improving production efficiency [5][7]. - Tencent's advertising tool, "Miao Si," exemplifies the integration of generative AI across various advertising processes, including content generation and cost reduction in distribution [7][8]. Group 3: Challenges and Opportunities in Generative Advertising - Traditional advertising recommendation systems face limitations, such as the difficulty in identifying user dislikes and the constraints of existing content libraries [9][10]. - A shift towards generative recommendation systems could address these issues by creating personalized content based on user behavior, although challenges remain in data availability and real-time processing [10][16]. Group 4: Tencent Advertising Algorithm Competition - The Tencent Advertising Algorithm Competition offers a platform for participants to engage with real business data, enhancing their understanding of user behavior and motivations [17][18]. - The competition features a total prize pool of 3.6 million RMB, with significant rewards for top teams, and serves as a recruitment avenue for Tencent [19][21]. - Participants gain valuable experience and networking opportunities, which can facilitate career advancement in the advertising technology sector [24][26]. Group 5: Market Trends and Future Prospects - Tencent's marketing services revenue grew by 20% year-on-year, largely attributed to AI-driven advertising technology upgrades, indicating a rising demand for generative AI talent in the industry [26][27]. - The competition encourages students from various academic backgrounds to participate, emphasizing that prior experience in advertising is not a prerequisite [28][29].

Core Insights - The AIGC industry in China is experiencing a significant early-stage investment trend, with total financing reaching billions of RMB in the first months of 2025, marking a 60% year-on-year increase [1] - Angel round financing events account for the highest proportion at 60%, indicating a preference for early-stage investments [3] Group 1: Current Situation - Early-stage projects have become the core area for capital allocation, with 60% of financing events occurring in the angel round, significantly higher than A rounds and strategic investments [3] - Startups established in 2025 account for 60% of the AIGC companies, with notable examples like "月之暗面" and "生数科技" completing significant financing within a year of establishment [4] Group 2: Driving Factors Behind Capital Preferences - Accelerated technological iteration is driving capital to focus on application-layer tools, allowing for quick validation of business models [6] - Policy support and market demand are also pushing the AIGC market, which is expected to exceed trillions by 2025, despite being only billions in 2025 [7] Group 3: Industry Participation - Major industry players like Tencent and Baidu are deeply involved in the ecosystem through strategic investments, with Tencent investing billions in 2025 [9] Group 4: Challenges and Pressures - Investors are increasingly demanding early-stage projects to demonstrate monetization pathways, with examples like "妙鸭相机" showcasing rapid customer acquisition through low-cost services [11] - There are signs of industry bubbles, with global AIGC financing exceeding hundreds of billions, but domestic projects facing challenges due to high levels of homogeneity [12] Group 5: Future Trends - Investment focus is shifting towards the middle layer of the industry, such as AI training tools and data annotation platforms, which are expected to enable scalable applications [15] - Global expansion is accelerating, with leading companies like "月之暗面" initiating overseas user growth plans, attracting capital interest in cross-language models and localization capabilities [15]

作者｜冬梅 字节 Top Seed 启动 2026 届招聘，瞄准顶尖博士 4 月 27 日，字节跳动 Seed 在其官微上发布了一则招聘启示，宣布正式启动 2026 届 Top Seed 大模型顶尖人才校招计划， 研究课题包括大语言模型、机器学习算法和系统、多模态生成、多模态理解、语音等方向，基本覆盖大模型研究各个领域， 计划招募约 30 位顶尖应届博士。 值得一提的是，本届 Top Seed 强调不限专业背景，更关注研究潜力，希望寻找具有极强技术信仰与热情、具备出色研究能 力、富有好奇心和驱动力的年轻研究者。 值得注意的是，字节跳动在此次招聘启事中还透露了几位刚毕业的同学已经做出了一些有影响力的研究。 比如，Z 同学构建并开源了首个多语言代码修复基准 Multi-SWE-bench，在 SWE-bench 基础上，首次覆盖 Python 之外的 Java、TypeScript、C、C++、Go、Rust 和 JavaScript 七种编程语言，1632 个真实修复任务，是真正面向"全栈工程"的评测 基准，其数据均来自 GitHub issue，历时近一年构建，以尽可能准确测评和提高大模型高阶编程智能水平。 ...