Welcome to the 【AI Daily】 column! Here is your guide to exploring the world of artificial intelligence every day. Each day, we present the latest hot topics in the AI field, focusing on developers to help you gain insights into technical trends and understand innovative AI product applications.
Fresh AI products click to learn more: https://top.aibase.com/
1. Alibaba's Tongyi Wanxiang Wan2.1-VACE Opensource: Claimed as the First Open-source Video Editing Unified Model
Tongyi Wanxiang announced the opening of VACE, supporting various resolutions and tasks, providing a one-stop video creation experience, and realizing efficient and flexible video editing through multimodal input mechanisms.
【AiBase Summary:】
✨ Supports text-to-video generation, image reference generation, local editing, and video extension, enhancing creation efficiency.
🌟 Powerful controllable rewriting ability, based on human pose and motion flow control, supports subject and background references.
🔧 Proposes video condition unit (VCU) for unified multimodal input, enabling free task combination and flexible editing.
Details link: https://github.com/Wan-Video/Wan2.1
2. OpenAI Upgrades ChatGPT: Officially Introduces GPT-4.1 with Superb Code Ability
OpenAI released GPT-4.1 and its lightweight version GPT-4.1mini, significantly enhancing coding capabilities and instruction execution experience while optimizing user experience and multimodal support, further consolidating its leading position in the AI field.
【AiBase Summary:】
🚀 GPT-4.1 has powerful coding capabilities, handling complex programming needs more efficiently with faster operation speed, making it an ideal choice for developers and instruction processing scenarios.
🌐 GPT-4.1mini is lightweight and highly efficient, still running smoothly on resource-constrained devices, providing broad access channels for both free and paid users.
🌟 ChatGPT adds new features such as long-press copy, table copy, and streaming transmission, significantly improving the user experience.
3. Stability AI Releases 341M Ultra-lightweight Text-to-Speech Model, Runs on Mobile Phones, Generates Audio in Only 8 Seconds!
Stability AI released an ultra-lightweight text-to-audio generation model named 'Adversarial Post-Training Accelerated Rapid Text-to-Audio Generation', with only 341M parameters, yet it can generate 12 seconds of audio in 75 milliseconds on H100 GPUs and complete the same task in 7 seconds on mobile CPUs, showcasing explosive performance and strong diversity.
【AiBase Summary:】
⚡️ The ARC post-training method, which does not rely on distillation, improves model generation speed and quality.
📱 Lightweight design, supports local mobile operation, greatly enhancing mobile creative application experience.
💫 Audio-to-audio function enables style transfer, inspiring more creativity.
Details link: https://arxiv.org/pdf/2505.08175
4. Poe Report: Keling Large Model Accounts for 30% of Generated Video Volume, Leading Runway
The recently released 2025 Spring AI Model Usage Trend Report shows that Keling’s multiple video generation models from Chinese Kuaishou perform outstandingly in the text-to-video domain, accounting for 30% of the market share. Among them, Keling 2.0 accounted for 21% of usage within three weeks of its release in April. Since its launch last June, global users have exceeded 22 million, with monthly active users increasing 25 times, generating significant numbers of videos and images.
【AiBase Summary:】
🌟 Keling large model accounts for 30% of the market share in the text-to-video domain, leading competitors like Runway.
📈 Keling 2.0 model accounted for 21% of video generation market within three weeks after its release in April.
👥 Global users of Keling AI exceed 22 million, monthly active users increase 25 times, and generated video and image numbers significantly increase.
5. Microsoft's WizardLM Team Joins Tencent, Possibly Integrated into Hunyuan Large Model R&D System
Microsoft's artificial intelligence research team, WizardLM, has joined Tencent AI Lab's "Hunyuan" team, marking Tencent's further efforts in the large model field. This team not only brought multiple technological breakthroughs but also demonstrated its R&D strength through open-source models.
【AiBase Summary:】
✨ The former Microsoft WizardLM team joins Tencent Hunyuan team, strengthening Tencent's competitiveness in the large model field.
🚀 Hunyuan-TurboS0416 model uses "Hunyuan" naming for the first time, symbolizing the deep integration of the team with Tencent.
💼 Tencent plans to significantly increase AI investment, aiming to occupy a more dominant position in the global AI competition.
6. Tencent Announces the Release of Hunyuan Image 2.0 on May 16th
Tencent's Hunyuan large model team announced that Hunyuan Image 2.0 will be released on May 16th, marking an important breakthrough for Tencent in the AI visual field, with the core concept of 'smarter, more open, more China-oriented'.
【AiBase Summary:】
🌟 Hunyuan Image 2.0 will be released on May 16th, marking another significant advancement for Tencent in the AI visual field.
🌐 New tools emphasize 'smarter, more open, more China-oriented', assisting creators and enterprises in entering the AI-driven visual production era.
🚀 Following last year's Hunyuan large model upgrade, Tencent once again demonstrates its continuous innovation power in the artificial intelligence field.
7. Shanghai Initiates Artificial Intelligence Identifier Ecosystem Alliance, Xiaohongshu and MiniMax Join as First Members
This article introduces the establishment of Shanghai's artificial intelligence identifier ecosystem alliance, which aims to promote the development of identifier technology in the artificial intelligence field, enhance transparency and security of generated content, and lay the foundation for building a trustworthy AI environment through policy interpretation and corporate cooperation.
【AiBase Summary:】
🌟 The alliance is guided by the Shanghai Cyberspace Administration and gathers many well-known enterprises, aiming to improve the transparency and security of AI-generated content.
🔍 The National Internet Emergency Center and China Electronics Standardization Research Institute interpret relevant policies, emphasizing the combination of international rules and Chinese characteristics.
🤝 Xiaohongshu, MiniMax and other companies participate in identifier work practices, explore various content identification solutions, and accumulate governance experiences.
8. Lightricks Releases LTX-Video-13B Refined Model! Generates High-Quality AI Videos in 10 Seconds, Speed and Quality Double Leap!
Lightricks, an Israeli tech company, released an open-source AI video generation model LTX-Video-13B refined model, which is based on 13 billion parameters and combines multi-scale rendering technology and efficient quantization optimization, boosting video generation speed to less than 10 seconds while maintaining high-quality output.
【AiBase Summary:】
🚀 Uses multi-scale rendering technology, generates high-definition videos in 10 seconds, speed increases more than 5 times.
🌍 Open-source model, supports low-memory device operation, reducing AI video production costs.
🌟 Generation speed improves 30 times, comparable to professional film works, reshaping the content creation ecosystem.
Details link: https://github.com/Lightricks/LTX-Video
9. Google AlphaEvolve Released! Gemini Self-evolving AI Solves Math Problems, Optimizes Chips and Data Centers, Training Speed Soars by 32.5%
Google DeepMind released AlphaEvolve, an AI coding agent combining Gemini large language model and evolutionary algorithms, demonstrating powerful self-optimization capabilities in multiple fields, including data center scheduling, chip design, AI training, and mathematical research.
【AiBase Summary:】
🌟 Combines Gemini with evolutionary algorithms, solving complex problems like chip optimization and math puzzles.
🚀 AlphaEvolve optimizes data center scheduling, recovering 0.7% of global computing power, saving operational costs.
🔍 Improves AI training efficiency; Gemini model training speed increases by 32.5%, showcasing strong self-optimization capabilities.
10. Tencent Yuanbao Browser Extension Beta Version Launched on Chrome
Tencent Yuanbao browser extension beta version is now available on the Chrome platform, offering features like floating ball, persistent sidebar, and word selection toolbar, improving web browsing and information processing efficiency.
【AiBase Summary:】
✨ Floating ball feature supports one-click translation and summary of web content, easily overcoming language barriers and saving reading time.
💬 Persistent sidebar can efficiently answer questions, supports screenshot questioning, greatly improving information acquisition efficiency.
🔍 Word selection toolbar allows instant search or translation after selecting text, making information processing smoother.
Details link: https://yuanbao.tencent.com/download