Welcome to the 【AI Daily】 section! This is your daily guide to exploring the world of artificial intelligence. Every day, we present the hottest topics in the AI field, focusing on developers to help you understand technology trends and discover innovative AI product applications.
Fresh AI products Click to learn more: https://top.aibase.com/
1. OpenAI announces ChatGPT search upgrade with support for maps, advanced voice features, and more
OpenAI announced a significant technical update for the ChatGPT platform during its latest live stream, introducing real-time search and advanced voice interaction modes that greatly enhance user experience. By optimizing the search algorithm, users can quickly access real-time information and view source links directly, improving convenience. Additionally, new video playback and map integration features provide a more intuitive search experience and enhance search efficiency on mobile devices.
【AiBase Highlights:】
📈 The updated ChatGPT introduces real-time search capabilities, optimizing the search algorithm to enable users to quickly obtain real-time information about stocks and news.
🗣️ The new advanced voice interaction mode allows users to conduct multi-turn searches via voice, offering a personalized voice assistant experience.
🗺️ ChatGPT now supports map integration, allowing users to directly view geographic information, plan routes, and explore locations.
2. Google upgrades AI video generation model Veo2, achieving 4K resolution and superior human preference ratings over Sora
Google recently released its next-generation video generation model, Veo2, aimed at competing with OpenAI's Sora. Veo2 demonstrates higher realism and detail in video generation, and users can apply for access through Google Labs' VideoFX platform. Additionally, Google has updated its image generation model, Imagen3, further enhancing the realism and color performance of generated images.
【AiBase Highlights:】
🎥 The Veo2 video generation model outperforms OpenAI's Sora, and users can apply for access.
🚀 Users can select video styles and effects, generating videos with resolutions up to 4K.
🎨 The updated Imagen3 image generation model showcases improved artistic styles and user experience.
Details link: https://labs.google/fx/tools/video-fx
3. Midjourney launches personalized model and mood board, allowing image uploads for model training
On December 16, 2024, Midjourney launched the highly anticipated "Mood Board" feature, allowing users to upload collections of inspirational images to generate new artworks. Coupled with the latest AI models, users can more easily create personalized profiles, simplifying the model-building process and lowering the entry barrier for new users. Additionally, enhanced organizational features enable users to better manage multiple projects.
【AiBase Highlights:】
🌟 Midjourney has launched the mood board feature, allowing users to upload collections of inspirational images.
🚀 Creating personalized profiles has become simpler, requiring only 40 ratings to get started.
🛠️ Enhanced organizational features allow users to name profiles and track related images.
Details link: https://www.midjourney.com/personalize
4. Google introduces new AI tool Whisk, allowing users to mix multiple images to generate new styled images without prompts
The latest AI tool from Google, Whisk, significantly changes traditional image generation methods, allowing users to upload multiple images to create new images without relying on lengthy text descriptions. Whisk is designed for rapid visual exploration, enabling users to easily blend images of different styles and themes to create unique visual works. Although the image generation process may take a few seconds and sometimes yield slightly odd results, the overall experience is very enjoyable.
【AiBase Highlights:】
🎨 Whisk allows users to generate new styled images by mixing multiple images, disrupting the traditional text prompt approach.
✨ Users can upload images of different themes, automatically blending them to create interesting visual effects.
🚀 Google has also released the Imagen3 and Veo2 models, further enhancing image and video generation capabilities.
Details link: https://top.aibase.com/tool/whisk
5. YouTube launches new feature allowing creators to authorize third parties to use their videos for AI training
YouTube has recently launched a new feature that allows creators to choose whether to authorize third-party companies to use their videos for training AI models. The default setting for this feature is off, so creators do not need to take any action if they do not wish for third parties to use their videos.
【AiBase Highlights:】
🔒 The default setting is off; creators must actively choose to allow third parties to use their videos for AI training.
🤝 Allowed third-party companies include well-known AI firms such as OpenAI, Apple, and Microsoft.
📈 This feature aims to help creators realize new value from their content in the AI era.
6. TuSimple releases its video generation model "Ruyi" and open-sources Ruyi-Mini-7B
Beijing-based TuSimple Technology Co., Ltd. released its first "image-to-video" large model "Ruyi" on December 17, 2024, and open-sourced the Ruyi-Mini-7B version for users to download and use on the Hugging Face platform. This model is designed for consumer-grade graphics cards and possesses various generation capabilities, especially showcasing outstanding visual storytelling potential in the anime and gaming sectors. Despite technological advancements, some defects still need to be addressed.
【AiBase Highlights:】
🚀 The Ruyi model is designed for consumer-grade graphics cards, supporting multi-resolution and multi-duration video generation, capable of handling resolutions from 384×384 to 1024×1024.
🎨 The model excels in frame consistency, motion fluidity, and color representation, making it an ideal creative partner for ACG enthusiasts.
🔧 Despite technological progress, Ruyi still has some flaws, such as hand deformities and facial detail loss, and TuSimple is working to improve these issues.
Details link: https://huggingface.co/IamCreateAI/Ruyi-Mini-7B
7. Zhiyun AI completes 3 billion yuan financing to promote large model technology research and commercialization
Zhiyun Company recently completed a new round of financing totaling 3 billion yuan, attracting participation from numerous strategic investors and state-owned enterprises. This funding will be used for the research and development upgrade of Zhiyun's base large model, further enhancing its capabilities in complex reasoning and multimodal task solving. Despite facing challenges from market competition and slowed technological progress, Zhiyun remains a leader in the AI industry and continues to have a significant impact globally.
【AiBase Highlights:】
🚀 Zhiyun has completed 3 billion yuan in financing, which will be used for the research and development and upgrade of its base large model, driving industry innovation.
📈 This year, Zhiyun achieved countercyclical growth in the B-end market, with API revenue increasing over 30 times year-on-year and the number of paying customers growing 20 times.
🌍 Zhiyun's C-end product "Zhiyun Qingyan" has attracted over 25 million users, and the expected paid features will generate tens of millions in revenue.
8. Meta launches open-source AI fitting model Leffa: retaining more details
Meta recently launched Leffa, an open-source AI virtual fitting framework aimed at enhancing users' dressing experience by generating new images. Users simply upload a reference image, and the system can generate new outfit effects, reducing the hassle of returns and exchanges due to poor fit. Leffa excels in retaining details and minimizing image distortion, providing a more natural fitting experience.
【AiBase Highlights:】
🌟 Leffa is an open-source virtual fitting framework launched by Meta that can generate new images based on reference images.
👗 The framework effectively reduces image distortion, retains more details, and enhances the virtual fitting experience.
💻 Users can try Leffa on the Hugging Face platform, and Meta provides complete project code.
Details link: https://github.com/franciszzj/Leffa
9. Diffusion-Vas: tracking video objects and completing occluded parts
In the field of video analysis, the persistence of objects is an important clue for understanding their existence. Researchers proposed the Diffusion-Vas method, based on diffusion priors, aiming to enhance the effectiveness of video unsupervised segmentation and content completion. This method consists of two stages: first generating unsupervised masks, and then using a conditional generation model to complete occluded areas. After multiple benchmark tests, this method has shown excellent performance in complex scenes, improving accuracy by 13%.
【AiBase Highlights:】
🌟 A new method has been proposed to achieve unsupervised segmentation and content completion in videos through diffusion priors.
🖼️ The method is divided into two stages: first generating unsupervised masks, followed by content completion of occluded areas.
📊 In multiple benchmark tests, this method significantly improved the accuracy of unsupervised segmentation, especially in complex scenes.
Details link: https://diffusion-vas.github.io/
10. Meta Ray-Ban Meta smart glasses upgrade: real-time AI video and translation features
Meta has made significant updates to its Ray-Ban Meta smart glasses, launching several new AI-based features, including real-time conversation and language translation. These features allow users to communicate with AI assistants more naturally without frequently waking them up, while also supporting instant translation between multiple languages, greatly enhancing communication convenience. Additionally, the glasses now feature Shazam functionality, enabling users to identify music through voice recognition.
【AiBase Highlights:】
🌟 The Ray-Ban Meta smart glasses introduce real-time AI video and translation features, allowing users to converse with AI assistants anytime.
🌍 The new real-time translation feature supports instant translation between multiple languages, enhancing communication convenience for users.
🎵 The glasses also support Shazam functionality, enabling users to recognize currently playing music through voice commands.
11. Broadcom CEO predicts explosive growth in AI market, company valuation surpasses $1 trillion
Broadcom CEO Hock Tan expressed an optimistic outlook on the AI chip market during a recent earnings call, predicting that Broadcom's revenue in this sector will grow significantly by 2027, with the addressable market estimated to be between $60 billion and $90 billion. The company's market capitalization has surpassed $1 trillion for the first time due to the surge in demand for AI chips.
【AiBase Highlights:】
🌟 Broadcom predicts that by 2027, the addressable market for AI will reach between $60 billion and $90 billion.
📈 Broadcom's market valuation has surpassed $1 trillion for the first time due to the soaring demand for AI chips.
💰 Through the acquisition of VMware, Broadcom's overall revenue has increased by 51%, with a significant reduction in operating costs.
12. Kingsoft Office: WPS AI to unlock four major features for free, including AI-generated PPTs
Kingsoft Office announced that WPS AI will provide four free features to users during the year-end period, aimed at enhancing work efficiency and creativity. Users can use AI to generate PPTs, clone styles, apply filters, and more to quickly create professional year-end summary PPTs. In addition, WPS offers a variety of PPT templates to meet the diverse needs of users.
【AiBase Highlights:】
🎉 WPS AI will unlock features for AI-generated PPTs, style cloning, filters, and templates for free, enhancing user work efficiency.
🖼️ The AI-generated PPT feature allows for quick creation of professional presentations, intelligently polishing content while maintaining logical and aesthetic design.
📋 Users can participate in the "AI Summary Season" activity to access a wealth of year-end summary PPT templates that meet various industry needs.
