Chinese video AI models expand globally

Chinese tech firms leverage the country’s vast video content market to deploy ads, film clips, and short-form videos automated by AI tools. Kling AI, developed by Kuaishou, a short video platform, ranks among the top players globally in converting text or still images into video content. 

According to Kuaishou, Kling AI is already used by more than 20,000 businesses, including advertisers and animation studios. At the recent World Artificial Intelligence Conference (WAIC) in Shanghai, Kling AI unveiled Kling 2.1, which can add sound effects to AI-generated visuals. Wei Xiong, Internet analyst at UBS Securities, revealed that AI video generation has the potential to reshape the content industry by enhancing production efficiency, reducing creative barriers, and unlocking new monetization models.

Chinese startups challenge Google’s video AI tools with top models 

TikTok’s parent company, ByteDance, dominates the AI video model space. Artificial Analysis, an AI research firm, ranks ByteDance tools first and third among the top five text-to-video models released in the last two months. Google models ranked 2nd and 4th, while Kling AI ranked fifth.  

Wei Xiong believes Chinese firms are still early leaders in the field despite growing international competition. He acknowledged current limitations with AI-generated video, including clip length, motion consistency, and user control over output. 

Zeng Yushen, Kling AI’s head of operations, said the Kling AI model isn’t limited to domestic users. She added that international markets like Japan, South Korea, and Europe are gradually becoming key focus areas. She told CNBC in an exclusive interview that they have observed that big AI models are increasingly globalized, so people don’t care about the country of origin the product is from. 

Kling AI generated over 150 million yuan ($20.8 million) in revenue in the first quarter of 2025, and daily spending on the platform reached 30 million yuan during the same period. The company hasn’t revealed Q2 results yet, but it confirmed that users can subscribe to the tool and purchase credits to create videos. Yushen declined to disclose the cost of training Kling’s AI models. 

Vidu, a rival startup from Beijing Shengshu, was launched globally about a year ago. By March 2025, it projected $20 million in annual revenue via subscriptions.

Chinese startups drive AI video innovation across multiple industries

Triolo, a partner at an advisory firm, noted that Chinese firms are applying the AI video tools across various industries. The startups identify pain points that companies are willing to pay to solve, leading to more effective AI applications. He cited 3DStyle, a Chinese startup that uses AI to design clothing styles and integrate them into automated, internet-connected manufacturing.  

Amazon, Google, and Microsoft also actively participate in the AI video tools industry. OpenAI unveiled its video generator Sora to ChatGPT users in December last year, almost a year after revealing the model. Kling AI went public in June 2024 ahead of Sora despite the early reveal. 

This week, Alibaba introduced Wan 2.2, the latest version of its video generation model. Wan 2.2 can control lighting, time of day, frame size, and camera angles. Wan is open-source and has been downloaded 5.4 million times since February from Hugging Face and China’s ModelScope platforms. 

Tencent recently launched Hunyuan World, which uses mesh-based 3D generation for game development. Daniel Ahmad from Niko partners revealed that the model supports Tencent’s game teams and reflects its goal of leading in game development. 

Niko partners revealed that over half of China’s game studios use AI to cut costs and development time. However, Ahmad warned that poor integration with AI-generated content may cause backlash in some cases.

KEY Difference Wire: the secret tool crypto projects use to get guaranteed media coverage