Overview
The latest SuperCLUE image-to-video (I2V) ranking places MiniMax H3 first domestically with a total score of 71.46 and an application score of 75.82. Alibaba’s Wan3.0 follows at 66.63; ByteDance’s Seedance2.5 and Kuaishou’s Kling V3Pro tie at 64.63 and 64.06. Overseas models are shown only as references: Google’s Gemini Omni1.1Flash at 67.38 and xAI’s Grok Imagine Video1.5 at 65.79.
Background & Interpretation
1. Technology Background & Evolution
Domestic AI video models have iterated rapidly over two years, moving from text-to-video to image-to-video, controllable generation and multimodal alignment. The ranking indicates domestic models have matched overseas peers overall and overtaken them on some application sub-items. Note that some models scored zero on tasks blocked by real-person limits, duration caps or content review, which affected totals.
2. Core Drivers & Underlying Mechanisms
Video-generation competition centers on consistency, duration, resolution and controllability. H3’s high application score reflects engineering gains in character/scene coherence, prompt adherence and physical plausibility. Meanwhile, inference efficiency is leaping (e.g., DeepSeek DeepInfer processing 155 trillion tokens/day on a single cluster) while training-side GPU idling persists—a core 2027 cost tension.
3. Market Structure & Industry Chain Impact
The ranking buoyed the domestic video ecosystem: model vendors, short-video and ad content toolchains, and cloud providers supplying inference all benefit. AI model photography in e-commerce, pre-visualization and game assets are accelerating, but distortion and disclosure-trust issues are emerging.
Implications & Outlook
Near term, video models will evolve toward longer duration, higher consistency and stronger control, converging with audio, 3D and world models. Watch: unified evaluation standards, falling compute cost, and content-labeling and copyright rules.
Takeaways for Industry Participants
Content producers can adopt cost-effective domestic video models; firms must build AI-content disclosure and review; vendors should close duration/consistency gaps and avoid compliance risk; investors can track the compute, tooling and rights chains video generation drives.
This column compiles industry information and shares technical perspectives; it does not constitute investment advice.