Financial news outlet reports that Orient Securities Company Limited has released a research note expressing optimism about the accelerated penetration of AI video in scenarios like short dramas, advertising, and professional filmmaking. The firm also sees AI gradually being applied in real-world industrial settings.

The significance of SD2.5, according to the brokerage, lies in two areas: first, it pushes filmmaking into a professional-grade realm, and second, it begins to explore industrial applications. Orient Securities Company Limited believes investors should focus on opportunities in vertical multimodal AI applications, predicting that technological breakthroughs and cost optimization will drive industry trends to accelerate, boosting user growth, payment penetration, and commercialization to new levels. The firm specifically highlights companies with leading multimodal technology and established commercial success.

Key Views from Orient Securities

MiniMax H3

Orient Securities Company Limited believes that MiniMax H3 and Seedance2.5 jointly raise the lower and upper limits of the AI video sector, with some differentiation in model positioning: SD2.5 raises the upper limit, expanding into high-value filmmaking and industrial scenarios, while H3 boosts the lower limit, filling the supply gap in the current competitive landscape with extreme cost-effectiveness and open-source availability. The firm sees AI video penetrating faster in areas like short dramas, advertising, and professional filmmaking, and gradually landing in industrial applications.

MiniMax H3: Raising the Lower Limit, Filling the Gap in Low-to-Mid-End, Cost-Effective Multimodal Supply

The H3 model's advantage lies in its native 2K output, providing image clarity that meets general creative needs. With a price of 0.8 yuan per second for 2K resolution, it offers strong value for money. Orient Securities Company Limited judges this could replace much of the demand for short dramas currently served by Seedance2.0. Moreover, H3 was officially open-sourced on August 2, allowing short drama companies to deploy the model independently, further reducing computing costs. Additionally, H3 offers differentiated performance in areas like advertising packaging and visual effects, which could accelerate AI adoption in ad production.

Seedance2.5: Breaking the Upper Limit, Accelerating Professional-Grade Filmmaking and Industrial Applications

Unlike Seedance2.0, which targeted broad users and the lower-end market, SD2.5's latest update focuses on breaking through the capability for high-quality, professional content creation, opening up new application scenarios and customer segments for AI video models. Key improvements include: (1) generating native 30-second clips, surpassing the previous industry limit of 15 seconds, enabling complete story arcs or single-shot long takes; (2) enhanced reference capabilities supporting up to 50 assets, avoiding the "uncanny valley" effect of identical faces in large crowd scenes; (3) better scene transitions and pacing tailored to filmmaking needs, plus timestamp-based video editing and 3D white-box pre-visualization features, all aimed at raising the ceiling for professional creation. In terms of pricing, the API price for SD2.5 at 720p is 42 yuan per million tokens, a 50% increase from SD2.0's 28 yuan for the same resolution. While this is a price hike, SD2.5 targets professional filmmaking, where labor and time costs are the primary expenses. Thus, even with the increase, the price remains acceptable. Overall, Orient Securities Company Limited concludes that SD2.5's significance lies in pushing filmmaking to a professional level and beginning to expand into industrial scenarios, such as generating engineering operation guidance videos, building autonomous driving training environments, and helping robots predict movement trajectories. By embedding video models into real production workflows, this could accelerate the application of AI in multiple industrial sectors, including manufacturing, embodied intelligence, and autonomous driving.

Risk Factors

AI technology iteration falling short of expectations, AI application deployment failing to meet expectations, and AI commercialization revenue generation failing to meet expectations.