AI・テクノロジー
"Kling 3.0" Officially Launched — 5-Minute, 4K Video Generation Reshapes the AI Video Production Landscape
機械翻訳 / Machine-translated
Kuaishou officially launched its video generation AI "Kling 3.0" on August 11, 2026. The model can generate up to five minutes of 4K-resolution footage from a single text prompt, while simultaneously achieving 2.8× faster generation speed, a 37% improvement in physical coherence scores, and a 38% reduction in API pricing compared to the previous generation. As of today, the industry's shared assumption that "video AI is still in the experimental stage" has been officially updated.
On August 11, Kuaishou simultaneously updated its official blog and API documentation to announce the general availability of Kling 3.0. The key specifications are as follows:
Within six hours of launch, responses from video creators spread rapidly, and "Kling 3.0" trended in Japanese-speaking communities as well.
"I generated a five-minute short film in one shot with Kling 3.0. The cuts were synchronized to the music. For product promo videos, it was at a level ready for real-world deployment starting today."
Competition in video generation AI accelerated sharply following OpenAI's Sora 2 (November 2025). Runway Gen-3 Alpha achieved up to 10 seconds at 1080p in June 2025, Pika 2.2 added lip-sync support in September of the same year, and Sora 2 reached 60 seconds at 4K — though it launched with a monthly generation cap still in place.
Kling 2.0 (March 2026) exploited that gap, capturing market share across Asia through a low-cost strategy despite being limited to two minutes at 1080p. Kling 3.0 is the first model to simultaneously achieve "long-form × high resolution × low cost" as its successor.
Kuaishou operates its own video platform with over 700 million monthly active users, and the vertical integration of content distribution infrastructure with model development represents a structural differentiator from competitors. The company has a setup in which large-scale data feedback from real-world services feeds directly into quality improvement.
The two-minute ceiling of previous models was sufficient for social media short-form video, but too short for corporate PR videos, music videos, and e-learning content. With five-minute support, part of the professional video market comes within range, shifting the competitive axis for production work.
Submission specifications for CTV (Connected TV) advertising and digital signage frequently require 4K. Kling 3.0 is seen as the first model to bring AI-generated footage up to "quality suitable for submission as advertising material," and is expected to reshape how production cost estimates are structured.
A price of roughly $1.50 per minute of video represents a two-order-of-magnitude difference compared to commissioning a professional videographer. For startups auto-generating product videos for e-commerce or demo videos for SaaS, this has crossed the threshold where the business model becomes viable.
Weaknesses of past generations — water flowing unnaturally, fingers breaking down — were markers that allowed instant identification of AI-generated footage. A 37% improvement on the VBench 2.0 score means a statistically significant increase in the proportion of scenes that general viewers cannot identify as AI-generated.
The Kling API is available in major countries including Japan, but the trajectory of U.S. government regulation of Chinese-made models in cloud AI remains a variable. Enterprises considering adoption have reached the point where they must evaluate data sovereignty and supply chain risk alongside cost-effectiveness.
What Kling 3.0 truly signals is not that "video AI has evolved," but that "the production economics of video content are being rewritten two to three years earlier than anticipated."
As of late 2024, there was an industry consensus that "video generation AI reaching practical levels would take until 2027 or later." Sora 2 moved that timeline forward to 2025, and Kling 3.0 has moved it forward again to August 2026. The direction in which predictions have consistently been wrong is uniformly toward "faster."
For the advertising, video production, and post-production industries, today is "the day the pool of procurement options expanded" — but simultaneously, "the day the basis for existing estimates began to erode." Whether one positions oneself as a user of Kling 3.0 as a tool, or remains in a position to be replaced by it, will likely be determined by how the next 90 days are navigated.
That said, the judgment involved in designing visual style, worldview, and brand personality remains human work. The more generative AI lowers the cost of "making," the more the value of "deciding what to make" rises in relative terms — and that structural dynamic will only intensify.
Kling 3.0 simultaneously satisfies four conditions — 4K, five minutes, low cost, and API availability — elevating AI video generation to a stage where it deserves serious evaluation as a procurement option in real business operations. The next points of focus are the new model Runway has previewed for autumn and beyond, and the trajectory of Sora 2's successor. The video generation AI market is likely to see further structural shifts within the next 60 to 90 days.
This article was written by an AI writer (AI News) from the Mirai News editorial team.
