Welcome to the 【AI Daily】column! This is your daily guide to exploring the world of artificial intelligence. Every day, we present you with the hottest AI news, focusing on developers and helping you understand technology trends and innovative AI product applications.
New AI Products Learn More: https://top.aibase.com/
1. Warning! Douyin Cracks Down on AI Stock Fraud, Targeting AI Stock Selection Tools and Courses
Douyin's safety center recently announced it will strengthen its supervision of stock investment content to create a healthy online environment. Some unqualified accounts have used AI tools to spread false information, misleading investors and committing fraud. Douyin has banned these accounts to protect user rights. Users should remain vigilant when investing to avoid being scammed.
【AiBase Summary:】
📉 Douyin encourages financial professionals to share authentic stock information and combat rumors.
🚫 Unqualified accounts are using AI tools for stock recommendations, misleading investors.
🔒 Douyin has banned violating accounts to protect user rights.
2. National Supercomputing Internet Platform Integrates Alibaba's Qwen Large Model, Providing QwQ-32B API
The National Supercomputing Internet Platform has partnered with Alibaba's Tongyi Qwen large language model to launch the Qwen QwQ-32B API service. This service offers developers and researchers free access to up to 1 million tokens, driving the development of AI technology. The Qwen QwQ-32B model has performed exceptionally well in several authoritative evaluations, demonstrating capabilities comparable to DeepSeek-R1 and becoming one of the most popular open-source large models on HuggingFace. Alibaba's Tongyi team continues to increase the number of open-source models, further solidifying its leading position in the global AI open-source field.
【AiBase Summary:】
🚀 The Qwen QwQ-32B API service is officially online, offering users free access to 1 million tokens.
📊 The Qwen QwQ-32B model performs exceptionally well in evaluations, comparable to DeepSeek-R1.
🌍 Alibaba's Tongyi team has open-sourced over 200 models, promoting the development of AI technology.
3. Kling Launches Three AI Special Effects: FuzzyFuzzy, MochiMochi, and BoomBoom!
Kling (Kling) has released its latest version, KLING 1.6, officially launching three AI special effects: FuzzyFuzzy, MochiMochi, and BoomBoom. These effects allow users to easily transform static images into dynamic videos, greatly enriching creative possibilities. FuzzyFuzzy, which transforms images into plush toys, has quickly gained popularity. MochiMochi offers a soft and bouncy visual experience, while BoomBoom attracts users with its energetic dynamic effects. The Kling team says this is just the first step in exploring AI effects and will launch more exciting features in the future.
【AiBase Summary:】
🎨 Three special effects, FuzzyFuzzy, MochiMochi, and BoomBoom, are now available. Users can experience AI magic by uploading images.
🧸 FuzzyFuzzy can transform images into cute plush toys with simple operations, making it a user favorite.
🚀 KLING 1.6, as the core technology support, provides powerful image-to-video generation capabilities, enriching user creative possibilities.
4. New Breakthrough in AI Digital Humans! Hedra Launches Character-3 Model and Hedra Studio: Joint Reasoning on Images, Text, and Audio
Hedra Studio recently launched the Character-3 model, marking a significant advancement in digital human video generation technology. Using multi-modal fusion technology, this model can process images, text, and audio to generate high-quality video content. Users simply upload their materials to quickly generate vivid virtual character videos with enhanced expressiveness and emotional control. Hedra Studio aims to lower the barrier to video production, enabling more creators to easily produce professional-level videos and foster creativity.
【AiBase Summary:】
🎥 The Character-3 model supports multi-modal input; users can generate virtual character videos by uploading images, text, or audio.
🌟 The new model features full-body motion capture and emotional control, enhancing the immersion and realism of video content.
🛠️ Hedra Studio, as a creative platform, aims to lower the barrier to video production, making it suitable for various users to quickly generate professional videos.
5. QQ Browser Launches AI Q&A Feature "Yuanbao Quick Answer"
On March 7th, QQ Browser launched an AI Q&A feature called "Yuanbao Quick Answer" to enhance user search experience. Based on Tencent HunYuan's "Fast Thinking" model Turbo S, combined with search enhancement technology, it can quickly retrieve real-time information from the entire network and extract key points, providing users with concise and clear answers. Whether simple or complex, users can quickly obtain refined answers and further view the original text or ask follow-up AI questions to expand their knowledge.
【AiBase Summary:】
💡 Uses Tencent HunYuan's "Fast Thinking" model Turbo S to improve search efficiency.
🔍 Yuanbao Quick Answer can quickly retrieve information from the entire network and automatically extract key points.
📚 Users can view the original cited text or ask follow-up AI questions for a deeper understanding.
6. Open-Source HeyGen is Here!
Heygem is an offline video synthesis tool designed for Windows systems that accurately clones a user's appearance and voice, ensuring user privacy and security. It uses text and voice to drive virtual avatars, utilizing advanced AI algorithms for high-precision appearance and voice capture. It supports multiple languages and natural language processing. Heygem has a user-friendly interface, making it easy for beginners to use while providing a secure creative environment to prevent data leaks.
【AiBase Summary:】
🌐 Heygem is an offline video synthesis tool that accurately clones a user's appearance and voice.
🗣️ It uses text and voice to drive virtual avatars and supports natural language processing and multiple languages.
💻 It has a user-friendly interface, is easy for beginners, and provides a secure and privacy-protected creative environment.
Details: https://github.com/GuijiAI/HeyGem.ai
7. Firecrawl Launches LLM.txt API: Generate LLM.txt for Any Website by Providing the URL
Firecrawl recently launched an LLM.txt generator API (Alpha version) to help users convert the content of any website into text files suitable for large language model (LLM) training. Users only need to provide the website URL, and the system will automatically crawl the content and generate text files in llms.txt and llms-full.txt formats for analysis and training. While in the Alpha stage, it offers asynchronous processing and status monitoring, allowing users to set the number of pages to crawl and whether to generate the full text.
【AiBase Summary:】
🌐 Users only need to provide a website URL to quickly generate text files suitable for LLMs.
📝 Generates two text formats: llms.txt for a concise summary and llms-full.txt for detailed content, meeting different needs.
🔒 Only supports public page processing. The Alpha stage limits the number of websites processed to 5000.
Details: https://docs.firecrawl.dev/features/alpha/llmstxt
8. AI Traffic Devourer: ChatGPT Ranks Among Top 10 Global Websites, But Shares Little Traffic
According to Similarweb data, ChatGPT reached 390.5 million visits in February 2025. Although month-over-month growth slowed, year-over-year growth was a remarkable 137%. ChatGPT ranked fifth globally, with 4 billion monthly visits, demonstrating its strong user base. However, despite its massive user base, outbound traffic is unusually low, with users rarely clicking on original content, raising industry concerns.
【AiBase Summary:】
📈 ChatGPT reached a record 390.5 million visits in February 2025, with year-over-year growth of 137%.
🌍 ChatGPT ranks fifth globally, with 4 billion monthly visits, accounting for 1.86% of global web traffic.
⚠️ Despite its large user base, outbound traffic is scarce, and users rarely click on original content, impacting information verification.
9. AI's "Sharp Eyes"! Finer-CAM Enables AI to Understand Images More Accurately and Classify More Clearly
The emergence of Finer-CAM technology significantly improves the capabilities of artificial intelligence in image recognition, allowing it to accurately identify subtle differences and overcome the ambiguity and uncertainty of traditional methods. By comparing similar categories, Finer-CAM can pinpoint unique features, providing clearer explanations and helping users better understand the AI's decision-making process. This technology not only surpasses traditional methods in accuracy but also achieves multi-modal zero-shot learning, showcasing its powerful application potential.
【AiBase Summary:】
🔍 Finer-CAM accurately identifies subtle differences in images by comparing similar categories, improving classification accuracy.
🖼️ This technology has powerful noise reduction capabilities, effectively removing background interference for clearer image interpretation.
📚 Finer-CAM supports multi-modal zero-shot learning, understanding text descriptions and accurately locating corresponding objects in images.
Details: https://github.com/Imageomics/Finer-CAM
10. Baidu's AI Creation App, Chengpian, Officially Integrates the Full-Blooded DeepSeek-R1
On March 10th, Chengpian, an application under Baidu Wenku, officially integrated the full-blooded DeepSeek-R1, marking a significant improvement in its AI-assisted functions. This upgrade enables Chengpian to more efficiently meet users' diverse needs in learning, life, and creation. Users can experience fast and accurate AI services through the intelligent assistant, greatly improving learning and life efficiency, becoming a "second brain" for users.
【AiBase Summary:】
🚀 After integrating DeepSeek-R1, Chengpian users can enjoy efficient AI services on App and PC web versions, supporting long-form text generation and in-depth editing.
📚 For students, Chengpian can quickly generate exam preparation plans, thesis outlines, etc., improving learning efficiency.
🔍 The upgraded Chengpian has deep thinking and full-network search capabilities, quickly providing solutions to help users cope with challenges in learning and life.
11. LanPaint, an Inference-Based Partial Repainting Method, Repairs Images Without Training
LanPaint is an image repair tool released by developer scraed on GitHub, aiming to provide high-quality image repair results without additional training. The tool achieves seamless repair by iteratively optimizing the model's denoising process. Users can easily integrate it using the standard ComfyUI KSampler process, lowering the barrier to entry. LanPaint also supports various complex repair tasks, demonstrating powerful image processing capabilities.
【AiBase Summary:】
🎨 Zero-training repair: Can be used immediately on any stable diffusion model without additional training.
🛠️ Easy integration: Uses the same workflow as the standard ComfyUI KSampler, lowering the barrier to entry.
🚀 High-quality repair: Provides high-quality, seamless image repair results and supports various complex repair tasks.
Details: https://github.com/scraed/LanPaint
12. In Just Four Weeks! Foxconn Launches Chinese Large Language Model FoxBrain
Hon Hai Research Institute under Foxconn has launched the new traditional Chinese large language model FoxBrain, marking significant progress in the Chinese AI field. Trained in just four weeks, this model demonstrates strong technical capabilities and aims to meet the high market demand for Chinese content generation and processing. FoxBrain supports text generation, translation, and Q&A, and will be released as open-source to promote the popularization and application of AI technology. Foxconn hopes to secure a place in the global AI competition with FoxBrain and will continue to optimize the model to address technical challenges.
【AiBase Summary:】
🌟 FoxBrain is a traditional Chinese large language model from Foxconn's Hon Hai Research Institute, trained in just four weeks.
📖 This model will be released as open-source to promote the development and popularization of Chinese AI technology.
🔍 FoxBrain can be applied to various scenarios such as text generation, translation, and Q&A, improving Chinese content processing capabilities.