Welcome to the "AI Daily" section! This is your daily guide to exploring the world of artificial intelligence. Every day, we present you with the latest hot topics in the AI field, focusing on developers to help you gain insights into technological trends and innovative AI product applications.

Fresh AI products Click to learn more: https://top.aibase.com/

1. Kunlun Wanwei Releases the First Open-Source Video Generation Model for AI Short Play Creation: SkyReels-V1

Kunlun Wanwei has announced the open-source release of the first video generation model for AI short play creation, SkyReels-V1, along with its controllable expression and motion algorithm, SkyReels-A1. This initiative aims to address the high costs and usability challenges of existing models. This innovation will advance the AI short play industry by providing enhanced capabilities for generating high-quality film and television content. SkyReels-V1 has been trained on high-quality data, achieving detailed control over character performances and supporting various combinations of expressions and actions, thereby enhancing the realism and expressiveness of the videos.

image.png

[AiBase Summary:]

🎥 SkyReels-V1 is China's first open-source AI short play video generation model, supporting 33 expressions and over 400 action combinations.

🤖 SkyReels-A1 is the first SOTA-level controllable expression and motion algorithm, capable of high-fidelity micro-expression restoration.

🌐 Kunlun Wanwei is committed to open-source technology, promoting innovation and development in the AI short play industry while reducing production costs.

Details link: https://github.com/SkyworkAI/SkyReels-V1

2. xAI Launches Grok 3 Model: Surpassing GPT-4o with 10x Computational Power

xAI released its latest flagship AI model, Grok3, on Monday, marking a significant advancement in the field of artificial intelligence. Grok3's computational power is ten times greater than its predecessor, Grok2, enabling it to analyze images and answer questions, surpassing OpenAI's GPT-4o in multiple benchmark tests. The new DeepSearch feature allows users to quickly access in-depth information analysis, further enhancing the practical applications of Grok.

image.png

[AiBase Summary:]

🧠 Grok3 is xAI's latest AI model, with computational power increased by 10 times compared to the previous version.

📈 Grok3 can analyze images and answer questions, outperforming GPT-4o in multiple benchmark tests.

🔍 The newly launched DeepSearch feature provides in-depth information analysis, enabling users to quickly access information.

3. Step Audio Releases a Dominant 130B Parameter Voice Model with Real-Time Conversation and Emotion Cloning

Step Audio has recently open-sourced a 130 billion parameter voice model, marking a significant breakthrough in voice interaction. This model combines voice understanding and generation control capabilities, showcasing robust multilingual support and dialect recognition, allowing for personalized voice interaction experiences. Additionally, its emotion control and voice cloning features offer users richer interaction possibilities, indicating that future voice AI will become even more intelligent and natural.

image.png

[AiBase Summary:]

🚀 This model is the industry's first real-time voice dialogue system that integrates voice understanding and generation control, featuring powerful control capabilities.

🌍 It supports multiple languages and dialects, including Chinese, English, Japanese, Cantonese, and Sichuan dialect, facilitating cross-language communication.

🎶 With voice cloning and emotion control features, users can freely set the emotional tone of the voice, creating personalized voice assistants.

Details link: https://github.com/stepfun-ai/Step-Audio/tree/main

4. Best Performance! Step Audio Releases Step-Video-T2V Video Generation Model

Today, Step Audio and Geely Automobile Group jointly launched two open-source multimodal large models, particularly the Step-Video-T2V video generation model, which demonstrates exceptional capabilities in video generation with its 30 billion parameters. This model excels not only in generating high-quality videos but also in accurately understanding instructions, supporting various camera movements, and producing realistic images that comply with physical laws. Developers can experience this model on the YUEWEN app and provide feedback to drive technological advancement.

image.png

[AiBase Summary:]

🌟 The Step-Video-T2V model has 30 billion parameters and generates 204 frames of 540P resolution video, leading globally in performance.

🎨 The model excels in motion smoothness and aesthetic quality, surpassing existing open-source video models.

📱 Both models are now available on the YUEWEN app, allowing developers to experience and provide suggestions to promote technological development.

Details link: https://github.com/stepfun-ai/Step-Audio

5. AI Prodigy Luo Fuli Begins New Job

Luo Fuli, a 95-born AI genius, has recently started a new career journey, although specific company details have not yet been disclosed. She gained widespread attention due to an invitation from Xiaomi's founder Lei Jun with a multi-million annual salary. Luo Fuli obtained a master's degree in computational linguistics from Peking University and has accumulated rich research experience at Alibaba's Damo Academy and DeepSeek.

image.png

[AiBase Summary:]

✨ Luo Fuli has started a new job, but the specific company has not been announced.

💰 Lei Jun invited Luo Fuli to join Xiaomi's team with a multi-million annual salary.

📈 The cost-effectiveness of the DeepSeek-V2 model has been described by Luo Fuli as the "king of cost-effectiveness."

6. Small Model + Large Model Collaboration! Mita AI Search Launches Shallow Research Feature

Mita Network Technology Co., Ltd. launched its new feature, "Shallow Research," on February 18, 2025, aiming to leverage AI technology to quickly gather and analyze information, enhancing research efficiency. This feature employs a collaborative architecture of small and large models, capable of searching and analyzing hundreds of web pages in a short time, significantly reducing research time. Although still in the shallow research stage, it can already provide preliminary analysis results, showcasing the tremendous potential of AI in information processing.

image.png

[AiBase Summary:]

🧠 The collaborative architecture of small and large models enhances research efficiency.

⏱️ Capable of completing the search and analysis of hundreds of web pages in 2 to 3 minutes.

📈 Future plans include adding capabilities for code execution and numerical analysis to deepen research.

7. HYBE's AI Voice Tool Supertone Play Launches: 10 Seconds for Voice Cloning

HYBE has launched a text-to-speech tool called "Supertone Play," aimed at providing creators with high-quality, expressive audio content. The tool supports English, Korean, and Japanese, with plans to expand to Spanish and Chinese in the future. Supertone Play features 150 voice characters for various creative needs and includes voice cloning capabilities, allowing users to create personalized characters based on their voices and adjust emotional tones.

image.png

[AiBase Summary:]

🎙️ Supertone Play supports multiple languages and provides 150 voice characters for creators to use.

🔒 The tool includes security measures such as voice verification codes and inaudible watermarks to prevent misuse of technology.

🌍 During the testing phase, the tool has been widely used in over 80 countries, with users actively downloading generated content.

Details link: https://play.supertone.ai/

8. Tencent E-Signature to Launch AI Contract Drafting Feature

Tencent's E-Signature public account recently announced the upcoming launch of an AI contract drafting feature, utilizing mixed-source and DeepSeek large model technologies to provide users with convenient contract generation and management services. With AI, Xiao Wang can easily generate rental contracts and add necessary clauses, while Xiao Li can quickly review potential risks in supplier contracts. Legal personnel like Lao Zhang can also efficiently manage contracts, avoiding information loss. This feature will greatly enhance the efficiency and accuracy of contract processing.

image.png

[AiBase Summary:]

🤖 AI can generate contracts with one click, simplifying the drafting process and reducing user operational difficulty.

🔍 AI serves as a risk scanning assistant, quickly identifying potential unfavorable clauses in contracts.

📂 AI enables intelligent contract management, helping users quickly find key information and avoid contract "disappearance."

9. Shenzhen Futian and Guangzhou Launch "AI Civil Servant": Deploying DeepSeek Large Model

Futian District in Shenzhen and Guangzhou have successively launched "AI Civil Servants" based on the DeepSeek large model, marking significant progress in smart government construction. The AI smart employee launched in Futian District has covered 240 business scenarios, significantly improving government service efficiency, with personalized customization generation time greatly shortened, and review time reduced by 90%. Guangzhou is also advancing the construction of a smart government system based on DeepSeek, demonstrating the potential of AI technology in optimizing government services.

image.png

[AiBase Summary:]

🚀 The AI smart employee launched in Futian District covers 240 business scenarios, improving government service efficiency.

⏱️ Custom generation time has been reduced from 5 days to minutes, with review time cut by 90%.

🌐 Guangzhou is building a smart government system based on DeepSeek, promoting the construction of a new digital engine.

10. NetEase Youdao Launches China's First AI Native Learning Hardware SpaceOne Integrated with DeepSeek-R1

NetEase Youdao launched a revolutionary learning hardware, the full-screen Q&A dictionary pen SpaceOne, on February 18, becoming China's first AI native educational hardware deeply integrated with large model capabilities. This product, centered around "draft paper-style reasoning," breaks through the limitations of traditional problem-solving tools, providing in-depth explanations across various subjects, including mathematics, physics, and chemistry, greatly enhancing learning efficiency.

image.png

[AiBase Summary:]

🖊️ SpaceOne integrates the DeepSeek-R1 reasoning model, simulating the teacher's thought process through visual step-by-step breakdowns of difficult problems, covering 10 major subjects.

📱 Featuring a sliding body and a 4.4-inch eye-protecting full screen, the vertical display efficiency is improved by 350%, optimizing the learning interaction experience.

🌐 With a built-in vocabulary of 78 million words and a mobile 4G module, it supports translation among hundreds of languages, breaking through the limitations of regional educational resources.

11. OpenAI Co-Founder’s Startup SSI Raises Over $1 Billion, Valued at Over $30 Billion

OpenAI co-founder Ilya Sutskever has raised over $1 billion for his new startup "Safe Superintelligence" (SSI), which is now valued at over $30 billion, making it one of the most valuable private tech companies in the world. SSI focuses on developing safe AI systems and has not yet generated revenue; its first product will be Safe Superintelligence. Sutskever emphasizes that the company will be completely insulated from external pressures, avoiding involvement in complex product competition.

image.png

[AiBase Summary:]

🌟 SSI is raising over $1 billion, with a valuation exceeding $30 billion.

🚀 The company focuses on developing safe AI systems and has not yet generated revenue.

🤖 Sutskever states that the company's first product will be Safe Superintelligence, while other business activities will be temporarily set aside.

12. OpenAI Considers Special Voting Rights to Guard Against Hostile Takeovers from Musk and Others

OpenAI is exploring ways to protect its control through special voting rights for its nonprofit board to fend off potential hostile takeovers, particularly from billionaire Elon Musk. This strategy aims to ensure that OpenAI maintains its core mission's independence after transforming into a for-profit entity.

image.png

[AiBase Summary:]

🌐 OpenAI is considering special voting rights to protect its board's control against potential hostile takeovers.

🤖 This move is primarily aimed at ensuring that the company retains its core mission's independence after transitioning to a for-profit entity.

💡 This strategy could impact the entire tech industry, prompting more companies to consider how to protect their control.