EB SECURITIES: AI applications are fully rolling out across the film and television sector; AI short comic dramas and films are launching together and are expected to maintain high growth.

EB SECURITIES has released a research report stating that the recent rally in the AI film and television sector has been driven by the resonance between industry events and underlying model capabilitiesnamely, Bona Film Group's first AI theatrical feature film securing a release date, combined with breakthroughs in GPT-6 Astra's multimodal capabilitiesprompting the market to reprice the commercialization potential of AI as it moves from short dramas toward medium- and long-form content. In 26H1, the domestic user base for AI drama and comic dramas is expected to exceed 600 million, and the combined number of AI drama/comic drama practitioners and individual creators has already surpassed 500,000. Looking ahead from the current juncture, 2027 will be a critical year for the industrialization of AI film and television and for batch trial production of long-form content. Market expansion will mainly come from two new tracksAI long-form dramas and AI theatrical filmswith technological iteration as the core variable. The main views of EB SECURITIES are as follows: The core catalysts for the sector's rally come from two main threads First, theatrical-grade AI feature films are reaching their landing point; second, general-purpose multimodal foundation models have unlocked the core bottleneck in long-narrative video generation. Bona Film Group's "Sanxingdui: Future Past" has been scheduled for nationwide release on October 23. The film runs 100 minutes and is the first domestic AI hyper-realistic theatrical feature film to obtain a public screening permit from the China Film Administration. The project was produced by Bona's AIGMS team using its self-developed Boka Film Cloud Studio, marking the official departure of AI film and television from the short-drama experimental scenario and its entry into the commercialization validation stage for theatrical films. The core significance of this project lies not in the box office of a single film, but in proving that AI-assisted full-chain production can meet the industrial standards for theatrical review and exhibition, providing the industry with a reusable production paradigm and opening up market expectations for AI long-form content. Second, on the technology front, the core upgrade of GPT-6 Astra lies in the deep integration of a native multimodal architecture with on-device inference efficiency. The model supports real-time joint parsing of text, images, audio, and video streams, reduces response latency by approximately 40% compared with the previous generation, and can perform localized processing of some complex tasks on mobile devices, reducing dependence on cloud computing power. In addition, the release of Seedance 2.5 at mid-year was likewise a milestone upgrade. The core change in Seedance 2.5 is that the model raises the single-generation video limit from 15 seconds to 30 seconds and can organize multiple logically connected shots within 30 seconds, enabling a story to unfold according to a complete structure of setup, progression, turning point, and conclusion. Market size: The current AI drama market is expanding rapidly and may maintain a high growth rate for a considerable period From the perspective of the current market size, AI film and television exhibits a structural characteristic of "short dramas leading the way, with long-form content beginning to show signs of explosive growth." In the domestic market, AI short dramas and comic dramas are the current core base. DataEye's "2026 H1 AI Drama and Comic Drama Data Report" shows that the market size is expected to exceed RMB 40 billion in 2026. In 26H1, the domestic user base for AI drama and comic dramas is expected to exceed 600 million, and the combined number of AI drama/comic drama practitioners and individual creators has already surpassed 500,000. ByteDance's Douyin and Hongguo Short Drama are developing well in the domestic market. According to QuestMobile data, as of July 2026, Hongguo Short Drama's DAU had reached 168 million, up 107% year-on-year. This figure has already exceeded the combined daily active users of the four traditional long-video platforms "iQIYI, Youku, Tencent Video, and Mango TV." On the user stickiness metric, in February 2026, Hongguo Short Drama's average daily time spent per user per month had reached 125 minutes. Unlike the "membership subscription + advertising" profit model relied on by iQIYI, Youku, Tencent Video, and Mango TV, Hongguo Short Drama adopts a "free viewing + advertising revenue sharing" model. Users can watch all content without paying any fees, and the platform monetizes by selling user attention to advertisers. This model completely eliminates the payment threshold on the user side. It should be noted that domestic market size statistics differ considerably in methodologythere are differences between the gross method and the net method, as well as between GMV and ARRso more attention should be paid to data trends, as horizontal comparisons are difficult. Looking ahead from the current juncture, 2027 will be a critical year for the industrialization of AI film and television and for batch trial production of long-form content. Market expansion will mainly come from two new tracksAI long-form dramas and AI theatrical filmswith technological iteration as the core variable. EB SECURITIES expects that trends in 2027 may include: 1) Production processes moving toward standardization and industrialization, with AI no longer merely generating assets but embedded across the entire workflow; 2) Content stratification, with low-cost standardized content serving the advertising and paid short-drama needs of traffic platforms, while IP migrates toward premium long-form content; 3) Continued acceleration of overseas expansion, with AI automatic redrawing, dubbing, and localization adaptation capabilities lowering the threshold for cross-language content production; 4) Seedance 2.5 supports single-generation output of up to 30 seconds, region-controllable editing, and multi-reference image input, adapting to professional film and television production needs, and Seedance 3.0 is highly likely to be a new-generation long-narrative video model and an enterprise-facing solution for the film and television industry. In addition, subsequent iterations of GPT-6, Kling, Vidu, and other video models will continue to optimize, jointly forming the technology-driven main line. Risk warnings AI technology development falls short of expectations, AI drama-related policies tighten, etc.