AI・テクノロジー
Runway Gen-4 Officially Launched — 4K/60fps Video Generation Is Reshaping the Division of Labor in Professional Film Production
機械翻訳 / Machine-translated
On August 1, 2026 (local time), Runway ML officially launched its video generation AI, "Gen-4." For the first time, the model supports 4K resolution and 60fps output, and the maximum generation length per clip has been extended from the previous 16 seconds to 60 seconds. Internal benchmarks indicate that temporal consistency has improved by 3.2× compared to "Gen-3 Alpha Turbo," and the numbers are beginning to suggest that the technology has reached a level of quality practical enough for production environments.
The key changes in Gen-4 are as follows:
Reports from video creators have been pouring in on X (formerly Twitter), including comments such as:
"I replaced all the B-roll in my short film with Gen-4. Some projects no longer need a green screen at all. My entire sense of what things cost has changed."
Since launching "Gen-2" in 2023, Runway has released a major update approximately every six months. "Gen-3 Alpha," released in the first half of 2025, reportedly secured multiple commercial contracts with film studios, and the company's IR announcements state that revenue from its enterprise plan grew 220% year-over-year compared to 2024.
At the same time, competition with Kuaishou's Kling 2.0 and Google's Veo 3 is intensifying. Kling 2.0 holds an advantage in physical simulation accuracy, while Veo 3 benefits from distribution integration with YouTube Shorts. Gen-4 has chosen to differentiate itself through "numbers that are visible on a spec sheet" — namely resolution and frame rate.
Under the previous 16-second limit, generating longer footage required stitching together multiple clips. The extension to 60 seconds means that the standard lengths for 30-second commercials and corporate promotions can now be covered in a single generation. Not only does this reduce the editing effort involved in cutting between clips, but it also eliminates the need to unify color tone and brightness across clip boundaries — a change with significant practical impact on production workflows.
The biggest challenge in video production has long been the "drift" problem, where the appearance of characters or objects breaks down between cuts. The score of 89 is based on internal evaluation and has not yet been verified by any external organization, but usage reports notably suggest that consistent representation of the same individual across multiple cuts has reached a practical level.
Rolling out to Pro plan users at no additional charge appears to be a strategy prioritizing the retention and lock-in of the existing user base. However, because high-resolution and long-form generation is designed to consume credits rapidly, heavy users will see their effective monthly costs rise. The Enterprise API price of $0.04 per second translates to roughly $1.20 per 30-second piece of commercial footage — a figure that is likely to fuel growing debate as a point of comparison against traditional location shoot costs.
The four workflows most likely to see early substitution are: B-roll shooting, portions of green screen compositing, mockup video production, and the conversion of storyboards into motion. In Japan, bottlenecks are expected around dependence on English-language prompts, rights clearance, and internal approval workflows, suggesting a lag of roughly 6 to 12 months before widespread adoption takes hold.
What stands out in this announcement is not the spec figures but the "architecture of the pricing." The API unit price of $0.04 per second is structured so that the more footage you generate at scale, the more easily costs flip in comparison to traditional production budgets. For video production companies, this is the moment when the calculation of "how many location shoots can we replace?" starts running in earnest.
That said, Gen-4's role remains limited to accelerating the generation of raw material. Directorial judgment, client approvals, and legal rights management still require human involvement. Whether the mere fact of using AI-generated footage becomes a brand risk is still a question with divided answers, depending entirely on how sensitive individual clients are to the issue.
More important than Runway's own trajectory is the question of when Kling and Veo 3 will release their competing updates. When all three players reach 4K/60fps parity simultaneously, the competitive axis is expected to shift away from specs and toward "ease of workflow integration" and "transparency in copyright handling."
Runway Gen-4 looks set to mark a milestone in the transition from "AI can make videos" to "AI can produce broadcast-quality footage." The next focal points will be benchmark evaluations by third-party organizations and the public disclosure of adoption cases in commercial projects. Before the answer arrives to the question of when "AI-generated video becomes indistinguishable from live action," the estimating standards used on production floors may already be starting to change.
This article was written by an AI writer (AI News) from the Mirai News editorial team.
