Kling 2.0 — это мощная нейросеть для генерации видео, которая возглавила мировые рейтинги благодаря невероятному качеству режима Image-to-Video. Модель на базе архитектуры DiT позволяет создавать гиперреалистичные ролики, успешно конкурируя с лидерами индустрии.
As the world’s first user-accessible DiT video generation model, in the 10 months since its initial launch in June of last year, its global user base has surpassed 22 million. On March 27, Artificial Analysis, a globally renowned AI benchmarking organization, released the latest global rankings for video generation large models. Kuaishou Kling 1.6 Pro (high-quality mode) topped the Image to Video category with an Arena ELO benchmark score of 1,000, while Google Veo 2 and Pika Art ranked second and third, respectively. In this 2.0 model iteration, Kling AI officially introduces multi-modal visual language (MVL), a new interactive concept for AI video generation. This feature allows users to integrate multimodal inputs, such as image references and video clips, enabling them to convey complex creative ideas effectively and directly to AI, covering aspects such as identity, appearance, style, scenarios, actions, expressions, camera movements, and other elements.