AI Daily: Douyin Launches Seedance 1.5 Pro; QQ Music Introduces Localized AI Composition; GPT-5.2-Codex Officially Unveiled

Welcome to the "AI Daily" column! This is your guide to exploring the world of artificial intelligence every day. Every day, we present you with the latest content in the AI field, focusing on developers to help you understand technical trends and learn about innovative AI product applications.

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

1. Doupo Launches Seedance 1.5 Pro, Directly Generate Audio-Visual Videos

The new audio-visual creation model Seedance 1.5 Pro from ByteDance has officially launched on Doupo, offering ordinary users a low-barrier experience for creating audio-visual videos. The model improves the entire process of text, image, and sound creation, achieving more natural speech synthesis, cinematic storytelling, and dynamic camera effects, allowing users to easily generate high-quality audio-visual content.

【AiBase Summary:】

🎥 Semantic + Audio-Visual Synchronization: The model can deeply understand the intent of the text and simultaneously generate matching images and sound effects, dialogue.

🗣 More Natural "Speaking": The generated character dialogue tone and rhythm are closer to real human expression.

🎬 Cinematic Storytelling + Dynamic Camera: The image comes with scene changes and camera movements, accurately capturing complex actions.

2. GPT-5.2-Codex Officially Launched: Overcoming Long-Range Coding Challenges with "Context Compression"

The release of GPT-5.2-Codex marks a major breakthrough in the field of intelligent code generation. It demonstrates excellent performance in code refactoring, system migration, and cybersecurity defense, while ensuring technical security through strict access control strategies.

【AiBase Summary:】

🧠 GPT-5.2-Codex enhances long-range task execution capabilities through context compression technology.

🛡️ Demonstrates professional-level capture-the-flag (CTF) practical skills in the field of cybersecurity.

🔒 Implements a multi-layer deployment strategy to ensure both technological accessibility and compliance with security standards.

3. Meituan Launches SOTA-Level Virtual Human Video Generation Model LongCat-Video-Avatar

Meituan's LongCat team has released the SOTA-level virtual human video generation model LongCat-Video-Avatar, which performs excellently in long video generation and has attracted widespread attention from developers.

【AiBase Summary:】

🎥 LongCat-Video-Avatar is a video generation model introduced by Meituan's LongCat team, supporting multiple tasks.

🧠 Uses Cross-Chunk Latent Stitching training strategy to improve the quality of long video generation.

📊 Achieves SOTA levels on multiple datasets, demonstrating strong application potential.

Details: https://github.com/meituan-longcat/LongCat-Video

4. Apple Releases New Multimodal AI Model UniGen 1.5, Integrating Image Understanding, Generation, and Editing

Apple's newly released multimodal AI model, UniGen 1.5, has made significant breakthroughs in the field of image processing. It integrates image understanding, generation, and editing functions, significantly improving work efficiency. Through the "editing instruction alignment" technology, the model can better capture user intent and enhance the accuracy of image editing. In multiple industry standard tests, UniGen 1.5 performed outstandingly, showcasing strong competitiveness.

【AiBase Summary:】

🧠 UniGen 1.5 integrates image understanding, generation, and editing functions, improving work efficiency.

🛠 Introduces "editing instruction alignment" technology, improving the accuracy of image editing.

📊 Performs well in multiple tests, showing strong competitiveness.

Details: https://arxiv.org/abs/2511.14760

5. ElevenLabs Integrates WhatsApp! AI Voice Assistant Can Answer Calls, Make Bulk Outbound Calls, and Revolutionize Customer Service

ElevenLabs announced on December 18, 2025, that its AI voice agent (Agents) has been deeply integrated with WhatsApp, enabling dual-mode interaction across all channels. This update not only improves enterprise customer service efficiency but also reduces deployment costs, significantly enhancing customer satisfaction and conversion rates.

【AiBase Summary:】

✅ Natural Call Handling: When a user calls the company's WhatsApp number, the AI responds with highly realistic voice.

✅ Active Outbound Services: Supports single or batch outbound calls, suitable for high-frequency scenarios such as appointment reminders, collection notifications, and customer follow-ups.

✅ Real-Time Voice Interaction: Whether the user sends a voice message or the AI makes an outbound call, it achieves low-latency, highly natural, and smooth conversations.

Details: https://elevenlabs.io/docs/agents-platform/whatsapp

6. ByteDance's AI Phone Cooperation Plan Exposed, No Token Sharing Policy Begins "Profit-Sharing for Access"

ByteDance is actively collaborating with several leading hardware manufacturers for AI phones, obtaining AI traffic entry points by pre-installing AIGC plugins and offering profit-sharing strategies to attract manufacturers to participate, thus enhancing its competitiveness in the AI field.

【AiBase Summary:】

📱 ByteDance collaborates with vivo, Lenovo, and Transsion to pre-install AIGC plugins to obtain AI traffic entry points.

💰 Uses a policy of no custom development fees and token sales sharing to attract mobile phone manufacturers to participate in traffic distribution and revenue sharing.

🤝 This collaboration aims to break through the resistance of mobile phone manufacturers against third-party AI and enhance AI competitiveness.

7. Claude AI Launches New "Task Mode": Smart Assistant Evolves, Can Ask Questions and Execute!

Claude AI's new task mode significantly improves the performance and user experience of smart assistants in complex tasks through innovative dual-panel design, structured action plans, and automated skill integration.

【AiBase Summary:】

🧠 Claude Task Mode introduces a dual-panel layout to improve task execution efficiency.

📝 Supports generating structured action plans and clarifying questions to ensure task accuracy.

🛠 Integrates various skills and model context protocols to achieve automated task processing.

8. Luma Releases Ray3 Modify: AI Video Editing Breakthrough, Preserves Actor Performance Details, One-Tap Clothing and Scene Changes

Luma AI released Ray3 Modify, achieving high-fidelity AI modifications of live-action videos, preserving actor performance details while supporting clothing and scene changes, bringing new creative possibilities to the film, advertising, and gaming industries.

【AiBase Summary:】

🎬 Achieve precise editing through character reference images and key frames.

🔄 Preserve actor performance details and support clothing and scene changes.

💰 Raised $900 million in funding and plans to build an AI computing cluster.

9. QQ Music Launches Localized AI Composition: Core Ultra Empowers, Generate Complete Songs in Minutes, Creative Privacy Not Leaked

QQ Music's localized AI composition function realizes complete local large model music generation through the Core Ultra processor, greatly reducing the creation threshold and ensuring privacy security.

【AiBase Summary:】

🎧 QQ Music launches a localized AI composition function, allowing users to create original songs without internet connection.

🔒 Uses a local reasoning path to protect creative privacy, data not uploaded to the server.

🧠 The Core Ultra processor includes an NPU, improving AI computing efficiency and extending battery life.

10. WeChat Yuanbao AI Assistant Adds New Features, Easily Set Reminders with One Sentence

WeChat Yuanbao AI Assistant now adds the feature of setting reminders with one sentence, allowing users to manage schedules and to-do lists more conveniently. At the same time, it has the ability to parse public account articles, images, and documents, and supports voice interaction, lowering the usage barrier and suitable for various user groups.

【AiBase Summary:】

📅 Yuanbao AI Assistant supports users setting reminders with one sentence, improving the convenience of schedule management.

🗣 Users can interact with Yuanbao through voice messages, convenient for elderly users.

📱 Yuanbao is not limited to reminder functions, but also provides practical functions such as article parsing and image recognition.