Baidu Intelligent Cloud and Shengshu Technology have officially launched Vidu, a groundbreaking video model, bringing new possibilities to the creative sector. Vidu is the first fully self-developed video model in China, with comprehensive performance benchmarks against OpenAI's Sora. It is designed for creative fields such as art design, game development, film and animation production, and social entertainment, aiming to inspire the productivity and creativity of the entire human race.
Vidu features two main functions: text-to-video and image-to-video, offering options for 4-second and 8-second video durations with resolutions up to 1080P. It maintains high dynamics, realism, and consistency while introducing unique capabilities such as character consistency, anime style, and text and special effect generation. Vidu's inference speed is fast; users only wait 30 seconds for a 4-second video clip, twice as fast as the industry's quickest, providing a nearly seamless creative experience.
Baidu Intelligent Cloud provides systematic, end-to-end large-model training services for Vidu, including extensive computing power support and full-chain guarantees. Vidu utilizes Baidu's Baigek AI heterogeneous computing platform, achieving "fast, stable, and cost-effective" large-model development with an effective training duration of 98.8%.
With the official launch of Vidu, Shengshu Technology and Baidu Intelligent Cloud look forward to advancing AI-enabled creative futures through technological innovation, unveiling more multimodal models for various applications, and leading us into a more efficient, convenient, and creative digital world.