Haitong Securities International·2025-10-17 09:11Investment Rating - The report does not explicitly provide an investment rating for the industry or specific companies involved in video generation technology. Core Insights - The video generation technology is transitioning from "concept demos" to "near-real-time creation," with significant advancements in speed and usability among leading models [10][11]. - Domestic models are rapidly closing the gap with international counterparts in terms of usability and image quality, shifting the competitive focus to compute reserves and data quality [13]. - The commercialization of compute-intensive AI models is becoming clearer, with tiered pricing for advanced features expected to be a standard practice [14]. Summary by Sections Event Overview - On October 16, 2025, Google released Veo 3.1, and OpenAI's Sora 2 launched on September 30, 2025, marking a new phase in short video generation and social distribution [10][11]. - All four models tested (Sora 2, Veo 3.1, Keling, and Jimeng) can generate a 5-second video in approximately 1-2 minutes [10][11]. Model Performance - Veo 3.1 excels in style reproduction and camera grammar, while Sora 2 offers the strongest photorealism but has limitations in clarity and landscape output [11][12]. - Keling and Jimeng demonstrate significant user-friendliness and are rapidly improving to match top international models [13]. Ecosystem and Competition - The gap between domestic and international model ecosystems is narrowing, with Chinese models showing notable competitiveness in usability and performance [13]. - The focus of competition is shifting from generational gaps in models to aspects like compute power and product refinement [13]. Commercialization and Economic Implications - The report highlights a trend towards tiered pricing for advanced features in AI models, driven by high-performance computing needs [14]. - The expected doubling of global data center electricity consumption by 2030 emphasizes the economic implications of AI inference on video generation services [14]. Implications for Film and TV Industry - AI video technology is expected to significantly reduce costs in various production stages, allowing for faster iterations from script to sample [15]. - The integration of AI tools like Veo 3.1 can compress production timelines, making the workflow more efficient and cost-effective [15].