Group 1 - The domestic GPU industry is experiencing a capital surge, with the "Four Little Dragons" of domestic GPUs—Biran Technology, Moore Threads, Muxi Co., and TianShu ZhiXin—actively pursuing capital market opportunities [1] - Biran Technology officially listed on the Hong Kong Stock Exchange on January 2, 2026, becoming the leading domestic GPU stock, with an opening price increase of 82% and a market capitalization exceeding 100 billion HKD [1] - The four leading domestic GPU companies are long-term strategic partners of SenseTime, collaborating based on a "computing power supply-algorithm demand" logic, aiming for a domestic substitution strategy [1] Group 2 - The successful listing of Biran Technology marks a significant moment for domestic GPU companies, highlighting the value of their collaboration with AI software vendors [2] - SenseTime has launched the first multi-episode generative AI agent, Seko2.0, which has successfully adapted to the domestic AI chip Cambricon, achieving a critical leap in domestic computing power across various AIGC core scenarios [2] - SenseTime has established a comprehensive domestic ecosystem by deeply adapting with over ten top chip companies, including Biran Technology and Cambricon, promoting the transition of domestic AI chips from "usable" to "user-friendly" [2] Group 3 - SenseTime has built a full-link cooperation system from computing power supply to scene implementation, collaborating with various domestic partners to provide high-performance domestic computing power at low costs [3] - The MTT S series GPU from Moore Threads has completed full adaptation with SenseTime's SenseCore large device, undergoing industrial-grade testing in training and inference of large models [3] - SenseTime has also launched the industry's first real-time video generation inference framework, LightX2V, which supports multiple domestic chips through a highly compatible plugin model [3] Group 4 - SenseTime's ability to act as a "connector" for domestic chip adaptation is rooted in its comprehensive technical capabilities, from computing power scheduling to application optimization [4] - The collaboration with Cambricon has led to significant enhancements in inference performance, with improvements exceeding three times through hardware-friendly innovations [4] - SenseTime's proprietary heterogeneous interconnection technology and scheduling platform have achieved unified training capabilities across various acceleration cards, significantly shortening training cycles [4] Group 5 - The technological advancements in domestic adaptation have translated into accessible product experiences, such as the Seko multi-episode generative AI agent showcasing efficient video creation capabilities [5] - Applications like Xiaohuanxiong are optimized for personal PCs and domestic terminals, providing secure and high-precision AI services [5] - New products like the Daxiao robot and RuYing marketing AI agent leverage the collaborative advantages with domestic chips to promote the domestic substitution process in various fields [6]

Core Viewpoint - The rise of domestic AI is marked by significant advancements in GPU technology, with companies like Moer Thread and Muxi achieving remarkable stock performance, indicating a growing recognition of domestic computing power in the AI sector [2] Group 1: Development of Domestic AI - Domestic computing power has evolved from a backup option to a comprehensive system encompassing hardware, models, and applications, laying the groundwork for future ecological development [2] - The release of SenseTime's Seko 2.0 and its compatibility with Cambricon's technology signifies a new phase in the domestic AI ecosystem, moving from mere compatibility to enhanced utility and collaborative applications [2][5] Group 2: Model Adaptation as a Foundation - Model adaptation is crucial for the domestic AI sector, with companies like SenseTime investing in hardware compatibility since 2020 to ensure effective model and computing power integration [4][5] - The successful adaptation of various domestic chips with SenseTime's SenseCore demonstrates the scalability and commercial viability of domestic computing power [5][6] Group 3: Application Development Challenges - The primary bottleneck for domestic AI lies in application development, as existing computing power and models have not yet translated into significant product performance improvements [10] - By 2025, it is projected that domestic AI computing power investments will reach 450 billion yuan, with a significant portion allocated for domestic chip validation and adaptation [11] Group 4: Future of Domestic AI Ecosystem - The domestic AI ecosystem is transitioning from a "backup" solution to a complete technological framework, requiring stronger hardware, model libraries, and community support [14][16] - The establishment of the "SenseTime Power Mall" aims to create a flexible product matrix for domestic AI, facilitating the integration and adaptation of various computing resources and services [15]

Core Insights - SenseTime officially launched the 2.0 version of its AI video generation tool, Seko, on December 15, positioning it as an integrated multi-episode generation tool aimed at short drama and comic creators [1] User Adoption - Since its launch in July 2025, Seko has surpassed 200,000 users, with short drama and comic creators accounting for 50% of the total user base [3] Technological Advancements - Seko 2.0 introduced SekoIDX technology to address character consistency in multi-episode generation, while the SekoTalk feature supports precise lip-syncing for two or more characters [3] - The tool can reduce traditional comic production cycles by 80% to 90% [3] - To lower generation costs, the company implemented a distillation technology called Phased DMD [3] - The open-source real-time video generation inference framework, LightX2V, has been adapted for domestic chips such as Cambrian and Muxi [3] Application and Collaborations - A real-life short drama generated by Seko, titled "Wanxin Ji," topped the Douyin AI short drama rankings [3] - SenseTime has formed a strategic partnership with Changjiang Film Group to explore the application of AIGC in film production, with plans to launch a series of short dramas and co-incubate theatrical films next year [3]

Group 1 - SenseTime Group Limited officially launched Seko2.0, the first integrated AI video generation tool designed for individual creators and studios in the short drama and comic drama sectors [1] - Seko1.0, launched in July 2025, quickly gained over 100,000 users within two months and surpassed 200,000 creators in less than six months, with 30% being short drama creators and 20% comic drama creators [1] - Seko2.0 features a new UI upgrade for enhanced visual effects, supports multi-episode generation with context management for up to 100 episodes, and includes intelligent scheduling for characters and scenes [1] Group 2 - The creation of AI short dramas and comic dramas involves generating a vast number of frames, with a single 5-second video requiring nearly 100,000 tokens, and generating 10-20 frames could demand 1,000,000 to 2,000,000 tokens [2] - SenseTime introduced Phased DMD distillation technology to significantly reduce the overall cost of multi-episode generation by combining staged distillation with expert mixture models [2] - The company has open-sourced LightX2V, the first real-time video generation inference framework, which can generate a 5-second video in under 5 seconds on consumer-grade graphics cards, with over 3.5 million downloads [2] Group 3 - SenseTime plans to offer domestic solutions for creators, aiming to initiate a "low-cost" era in AI video creation with exceptional cost-performance [3]

Group 1 - The AI short drama and comic drama sector is experiencing explosive growth, with most current AI video generation products only supporting single short episodes or clips [1] - SenseTime officially launched its product release week on December 15, introducing several innovative and practical AI products, including the core product Seko 2.0, which is the industry's first integrated multi-episode generation intelligent agent [1] - Seko IDX addresses the challenge of character consistency across multiple episodes and scenes by introducing "negative reference images" during the high-noise phase of the diffusion model, ensuring stability while responding to various expressions, postures, and scene commands [1] Group 2 - Traditional workflows require over three months for a team to complete 50 episodes, but with Seko 2.0, the production cycle can be shortened by 80% to 90% [1] - SenseTime is tackling the high token demand and cost issues in AI short and comic drama multi-episode creation, where a 5-second video requires nearly 100,000 tokens, and the total token demand for 10-20 scenes can reach millions [1] - SenseTime has made two core technological breakthroughs to address cost challenges: the innovative Phased DMD distillation technology and the open-source real-time video generation inference framework LightX2V, which can generate a 5-second video in under 5 seconds on consumer-grade graphics cards [2]

Core Insights - The article discusses the competitive landscape of video generation models, highlighting the advancements made by various tech companies, including Google, Runway, and Kuaishou, while questioning the readiness of these models as productivity tools [2][9] - SenseTime's Seko 2.0 version is introduced as a significant advancement, enabling AI short drama creation with minimal human input, effectively allowing a single person to manage the production [2][4][7] Group 1: Industry Developments - Major tech companies are racing to release enhanced versions of video generation models before the end of the year, with Google launching Veo 3.1 and Runway introducing Gen-4.5 [2] - SenseTime's Seko 2.0 has been successfully deployed in over a hundred short drama studios, showcasing its capability to generate scripts, storyboards, and videos rapidly [7][9] Group 2: Technical Challenges - The article outlines the "impossible triangle" of video generation, where efficiency, cost, and quality are at odds, making it difficult for AI video generation models to meet commercial demands [11][13] - Current models, even at the Sora 2 level, require several minutes to generate just 10 seconds of video, which hampers rapid iteration and real-time feedback essential for industrial production [11][12] Group 3: Innovations in Video Generation - SenseTime's LightX2V framework is highlighted as a breakthrough in real-time video generation, achieving generation times of under 5 seconds for 5-second videos, significantly faster than current industry standards [16][17] - The framework employs Phased DMD technology, which enhances video quality and consistency while maintaining high generation speeds [19][20] Group 4: Engineering and Optimization - LightX2V incorporates a comprehensive optimization strategy across five dimensions: model, scheduling, computation, storage, and communication, enabling low-cost and real-time video generation [31][32] - The framework's architecture allows for efficient use of consumer-grade GPUs, achieving real-time generation capabilities with a memory requirement of less than 8GB [36][37] Group 5: Domestic Chip Adaptation - SenseTime's Seko 2.0 has achieved full compatibility with domestic AI chips, allowing for a cost-effective alternative to NVIDIA chips while maintaining comparable video quality [39][40] - The strategic support for domestic AI ecosystems is emphasized, marking a significant step for China's AI industry in achieving core technological independence [42]

Core Viewpoint - The LightX2V project has gained significant popularity in the ComfyUI community, achieving over 1.7 million downloads in a single month, enabling creators to generate high-quality videos in real-time on consumer-grade graphics cards [2][7]. Group 1: Technology and Performance - LightX2V utilizes a comprehensive inference technology stack aimed at low-cost, high-real-time video generation, achieving near 1:1 real-time video generation [2][7]. - The project features a dual-core algorithm: Phased DMD step distillation and LightVAE, which work together to compress the video diffusion process from 40-50 steps to just 4 steps while maintaining time consistency and motion details [10][11]. - LightVAE is designed to meet the dual demands of throughput and resolution in video generation, effectively reducing encoding and decoding overhead while maintaining high-quality visuals [12]. Group 2: System Optimization - After algorithmic compression, LightX2V employs a full-stack inference framework to enhance performance, making it efficient for both single-card and multi-card deployments [14][16]. - Key technologies include low-bit operators, sparse attention, and feature caching, which collectively reduce memory requirements to below 8GB, allowing entry-level consumer cards to run the system [21]. Group 3: Ecosystem and Applications - LightX2V supports a range of mainstream video generation models and is integrated with ComfyUI, allowing users to easily access accelerated inference through a familiar graphical interface [19][21]. - The project caters to various user needs, from individual creators to enterprise-level applications, enabling functionalities such as image-to-video and text-to-video generation [19][21]. - LightX2V is compatible with a variety of hardware, including both NVIDIA and domestic AI chips, facilitating localized and large-scale deployments [21].