Welcome to the [AI Daily] section! This is your daily guide to exploring the world of artificial intelligence, where we present the hottest content in the AI field every day, focusing on developers, helping you understand technical trends and innovative AI product applications.
Fresh AI products click to learn: https://top.aibase.com/
1. Zhipu AI's Full Model Matrix Price Reduction, Qingyan Intelligent Agent Center Launches "AI Lao Luo" by Luo Yonghao
Zhipu AI announced a significant price reduction for multiple models at the Zhipu AI Open Day, with its Qingyan App boasting over 300,000 active intelligent agents and a daily Tokens call volume of up to 400 billion times. Luo Yonghao's AI agent has officially joined the Zhipu Qingyan App Intelligent Agent Center, with the job number 001, open to society.
【AiBase Summary:】
🚀 Zhipu AI announced that the prices of the GLM-4-Air and GLM-3-Turbo models have been reduced to 0.6 yuan/million Tokens, and the Embedding-2 model debuts at 0.3 yuan/million Tokens.
💡 The GLM-4-Flash model has the highest price reduction to 0.06 yuan/million Tokens.
📈 Qingyan App has over 300,000 active intelligent agents, with a daily Tokens call volume of up to 400 billion times.
Details link: https://top.aibase.com/tool/zhipuqingyan
2. ComfyUI Time Node System TimeUi: Quick Add, Delete, or Rearrange Content
This article introduces TimeUi, a ComfyUI time node system similar to a video editing tool timeline interface. Users can quickly add, delete, or rearrange content, providing a smooth user experience. TimeUi offers convenient image upload and management functions, supports custom settings and various parameter adjustments, and supports different time formats and duration units, as well as Bezier curve features, providing users with a more powerful and flexible timeline node system.
【AiBase Summary:】
⭐ Users can directly upload images to nodes or attach other "upload image" nodes, simplifying the workflow.
⭐ Each timeline row includes various custom settings, such as toggling image mask visibility, easily fine-tuning the output.
⭐ You can change the timeline's time format and animation duration, choose between frames or seconds, and support Bezier curve features.
Product entry: https://top.aibase.com/tool/timeui-a-comfyui-timeline-node
Details: https://www.chinaz.com/2024/0605/1621434.shtml
3. Backseat AI: The Free AI Game Companion for League of Legends
Backseat AI is a free AI companion approved by Riot Games, designed for League of Legends players, aiming to help players improve their game performance. Players can get real-time game suggestions, including purchase recommendations and lane strategy, through voice overlay and in-game commentary. The AI companion does not consume excessive system resources and supports creators' creations and developments.
【AiBase Summary:】
🎮 Provides real-time game commentary and suggestions, including purchase recommendations and lane strategy.
🤖 Main features include Backseat Buddy providing real-time commentary and insights, post-match lobby viewing match summaries, pre-match suggestions helping players prepare.
💡 Future versions will introduce real-time Q&A features and various creator voice packs, offering more language support and personalized experiences.
Details link: https://top.aibase.com/tool/backseat-ai
4. DeepSeek Announces MiniCPM Free Commercial Use
DeepSeek announces the free commercial use of MiniCPM, demonstrating its faith and passion for the open-source community. MiniCPM-Llama3-V2.5 is a special gift for the open-source community, featuring a multi-modal model on the edge, garnering global attention. It is particularly suitable for visually impaired individuals, offering powerful OCR capabilities to improve document reading experiences.
【AiBase Summary:】
🎁 MiniCPM-Llama3-V2.5 is an 8B-sized edge multi-modal model, surpassing Gemini Pro and GPT-4V, supporting over 30 languages.
💼 MiniCPM and MiniCPM-V models are open for academic research and allow commercial use, subject to the Apache2.0 and MiniCPM Model Community License Agreement.
🚀 MiniCPM-Llama3-V2.5 has SOTA OCR capabilities, efficiently running on mobile devices, with image encoding speeds 150 times faster.
Details link: https://modelbest.feishu.cn/share/base/form/shrcnpV5ZT9EJ6xYjh3Kx0J6v8g
5. Coze Bots User Usage Data Analysis
In this article, X blogger @eviljer shares a presentation of Coze Bots data analysis and visualization charts, highlighting the highest proportion of games and education categories. Through data analysis, you can intuitively understand the popularity and development of various categories.
【AiBase Summary:】
📊 Games and education categories have the highest proportion, at 20.6% and 19.2% respectively, with the character category also performing well, accounting for 11.9%.
📈 The compactness of the data distribution shows that lifestyle, games, and efficiency tool categories are leading, while learning and education categories have room for development.
📉 Coze Bots top ten category data performance includes games, learning education, public configuration, characters, efficiency, text creation, image audio-video, lifestyle, business services, and code assistants.
6. True Fit Utilizes Generative AI to Help Online Shoppers Find Well-Fitting Clothes
True Fit launches a new feature "Fit Hub" to improve online shoppers' ability to find clothes that fit their body type using generative AI, integrating product page information to help shoppers quickly understand products and determine the right size, reducing return rates. It plans to introduce more features such as "Shopper Insights" and "Brand Sizing" to provide a deeper personalized shopping experience.
【AiBase Summary:】
⭐️ True Fit launches a new feature "Fit Hub" to improve online shoppers' ability to find clothes that fit their body type using generative AI.
⭐️ Fit Hub integrates product page information, helping shoppers quickly understand products and determine the right size, reducing return rates.
⭐️ Plans to introduce more features such as "Shopper Insights" and "Brand Sizing" to provide a deeper personalized shopping experience.
7. Apple to Integrate AI Technology in iOS18 at WWDC
Apple's upcoming iOS18, to be unveiled at WWDC, will fully integrate AI technology, bringing a series of anticipated new features and improvements. This update covers AI support for Siri, Apple Photos, voice recording, and Notes applications, AI updates for iMessages, and other AI features, demonstrating Apple's commitment to integrating AI technology into every aspect of the operating system.
【AiBase Summary:】
🌟 Siri Upgrade: Siri will be able to control in-app functions, launch on Apple Watch, have a more natural and fluent voice, and have the ability to summarize notifications.
📸 Apple Photos New Features: A complete interface update, introducing a "cleanup" function, using AI to retouch photos, and internally developing a Generative Playground application.
🔊 AI Support for Voice Recording and Notes Applications: AI to summarize notes, transcribe audio content into notes, Math Notes to help create graphs and solve equations.
8. CamCo: 3D Consistent Image-to-Video Generation with Camera Control
This article introduces a novel image-to-video generation framework called CamCo, which achieves 3D consistency by introducing Plücker coordinates and epipolar constraint attention, and improves the synthesis effect of object motion by fine-tuning real-world videos with structured light algorithms. CamCo significantly improves camera control capabilities and 3D consistency, capable of generating high-quality, reasonable object motion videos.
【AiBase Summary:】
🔍 CamCo is a 3D consistent image-to-video generation framework with camera control, introducing Plücker coordinates and epipolar constraint attention to achieve 3D consistency.
🔍 Integrates epipolar constraint attention in each attention module, fine-tunes real-world videos with structured light algorithms to improve the synthesis effect of object motion.
🔍 CamCo processes real-world videos through a data screening process to enhance the ability to generate videos with camera self-motion and dynamic subjects.
Details link: https://top.aibase.com/tool/camco
9. Nvidia's RTX Remix Modders Introduce Comfyui Compatible Nodes
Nvidia's RTX Remix modders have introduced new nodes compatible with Comfyui, bringing a significant advancement to game modding and resource enhancement. Modders can now upscale and redraw game assets within the Comfyui interface, simplifying the editing and upgrading process. This initiative provides super-resolution and PBR models, improving the visual effects of games and enhancing the immersive experience for players.
【AiBase Summary:】
🎮 Modders can directly upscale and redraw game assets within the Comfyui interface, simplifying the editing and upgrading process.
🔧 Using the new nodes, modders can export game textures from RTX Remix to Comfyui and enhance them in batches through AI models.
🌟 The RTX Remix Toolkit is equipped with a powerful REST API, helping modders modernize remaking DirectX8 and DirectX9 games, driving innovation and development in RTX remaking projects.
Details link: https://www.nvidia.com/en-us/geforce/news/rtx-remix-rest-api-comfyui-app-connectors/
10. Zhipu AI Announces Open Source GLM-4-9B Model
The Zhipu AI team open-sourced the GLM-4-9B model on March 14, 2023, attracting widespread attention. The model introduced large language models for data filtering in pre-training, increasing training efficiency by 3.5 times. GLM-4-9B has more powerful inference performance, longer context processing capabilities, multilingual, multimodal, and All Tools capabilities.
【AiBase Summary:】
🚀 GLM-4-9B model open-sourced, attracting widespread attention
💡 Pre-training adopted large language models, training efficiency increased by 3.5 times
🔥 GLM-4-9B has powerful inference performance, multilingual, multimodal, and All Tools capabilities
Details link: https://github.com/THUDM/GLM-4
11. Musk's Grok Now Accessible on xAI Website
Musk's Grok is now accessible on the xAI website, with this tool offering more features on the xAI platform than the previous version showcased on Twitter. Notably, its branching tree mode allows users to observe different answer paths that may arise in multi-turn dialogues, which is very useful for analyzing and understanding dialogue flows. The Grok-1.5Vision model has demonstrated outstanding performance in multiple benchmark tests, surpassing the industry-leading GPT4V model, especially in the RealWorldQA benchmark test.
【AiBase Summary:】
🔍 Grok is now accessible on the xAI website, with more features
🌳 Branching tree mode allows observation of different answer paths in multi-turn dialogues