Soochow: Synergy Between Model Capabilities and Production Workflows Brings AI Film and Television Industrialization into a Validation Phase

Soochow released a research report stating that it is bullish on AI film and television moving from technology demonstration to commercial delivery, and that improvements in model capabilities and coordination with production workflows are expected to drive content supply expansion and improve production efficiency. It is recommended to pay attention to relevant companies with high-quality IP reserves, content production capabilities, or AI tool platform layouts, and to seize the development opportunities brought by industrialization validation. Soochow's main views are as follows: AI film and television is moving toward industrialization, with short-form content validating first and feature-length production accelerating AI dramas and animated dramas, relying on low cost, rapid iteration, and serialized production, have taken the lead in forming content supply, with the industry extending from short videos to films, series, and theatrical releases. Bona Film Group's "Sanxingdui: Future Past" will hit theaters nationwide on October 23, marking the stage when AI content begins to enter big-screen delivery. The bank believes that industry competition will shift from single-model capability to coordination among technology, content, and production workflows, with the core lying in high-quality IP and large-scale delivery capability. AI film and television is moving toward industrialization, with short-form content validating first and feature-length production accelerating Model competition is shifting from general generation to scenario differentiation, with Utopai entering the top tier: according to the latest Artificial Analysis text-to-video with audio rankings, Wan 3.0, Utopai X, Seedance 2.5, and MiniMax H3 rank in the top four respectively. Wan 3.0 places greater emphasis on all-around multimodal production, supports text, image, video, audio, and document inputs, can generate videos up to 30 seconds long, and strengthens native audio, reference consistency, and video editing capabilities; MiniMax H3 adopts a 33B-parameter Dense Single-Stream Transformer, organizing text, visual, and audio information into a unified sequence, with a focus on improving coordinated audio-video generation capabilities; Seedance 2.5 adopts a unified multimodal audio-video joint generation architecture, highlighting multi-material references, multi-shot narrative, and timestamp-level editing. Utopai X uses MiniMax H3 as its base for post-training in film and television scenarios, carrying out targeted optimization in audio-visual synchronization, physical performance, and camera control, and connects the model to script, character, scene, and version management through PAI. The bank believes that leading models have already formed different capability emphases, and Utopai's differentiation lies not only in ranking second, but more in combining the base model, film and television post-training, and production workflows, pushing competition from "generating a single good shot" to "completing a film that can be continuously revised and delivered." AI film and television is moving toward industrialization, with short-form content validating first and feature-length production accelerating Commercialization is moving from tool services to content assets: the industry currently has two main paths: first, the tool business, providing video models and production tools to creators and film and television institutions through subscriptions, API calls, or charges based on generated duration; second, the content business, in which companies bear project development, financing, and production risks, and then recoup investment through copyright pre-sales, television and streaming broadcast rights, theatrical distribution, and subsequent IP licensing. Utopai is an AI-native film and television company developed from Cybever, with simultaneous layouts in the Utopai X video model, the PAI production platform, and film and television project development and production. Its core approach is to embed models into script breakdown, character and scene management, shot generation, version iteration, and post-production collaboration workflows, improving the continuity and deliverability of AI content, thereby breaking through the obstacle that AI works find it difficult to enter traditional procurement systems. Utopai has already pre-sold part of the international distribution rights for "Space Nation" and "Corts" to buyers such as Globo TV and ZDF Studios, at market prices comparable to non-AI content. The bank believes that the tool model can generate cash flow relatively quickly, while the content model has a higher revenue ceiling and IP value. In the future, the industry may form a combined model in which "tool services provide basic revenue and content production amplifies commercial returns." Risk warning: AI progress falling short of expectations, industry competition exceeding expectations, and theatrical consumption falling short of expectations.