Feng Huang Wang·2025-12-16 06:22Core Insights - Alibaba's Tongyi Wanxiang team has launched the new Wanxiang 2.6 model, which is the first video generation model in China to support role-playing features [1] - The model integrates capabilities such as audio-visual synchronization, multi-shot generation, and sound-driven functionalities, aiming for overall consistency in generated videos [1] - The upgrade enhances video quality, sound effects, and instruction adherence, allowing for video generation of up to 15 seconds in length [1] Technical Features - The Wanxiang 2.6 model employs multi-modal joint modeling to learn temporal information, subject characteristics, and acoustic elements from input videos [1] - The storyboard control feature can construct professional narrative segments with multiple shot transitions based on semantic understanding [1] - Users can upload personal videos and use prompts to automatically design storyboards, perform role-playing, and provide voiceovers, creating cinematic short films [1] Target Applications - The new capabilities are primarily aimed at professional scenarios such as advertising design and short drama production [1] - The Wanxiang model family now includes over ten visual creation capabilities, such as text-to-image, image editing, and text-to-video [1] - Users can experience Wanxiang 2.6 through the official website, and enterprise users can access the model API via Alibaba Cloud's Bailian platform [1]