AI・テクノロジー
Runway "Gen-4" Officially Launched — Up to 240 Seconds of Consistent Footage and 6-Axis Camera Control Transform the "Shooting Phase" of Video Production
機械翻訳 / Machine-translated
On September 10, 2026, Runway simultaneously released its new flagship video-generation AI, "Gen-4," for both the API and Runway Studio. It comes standard with temporal consistency of up to 240 seconds (4 minutes) and "Cine Control," a 6-degrees-of-freedom camera path system. The API price of $0.06/second represents approximately a 40% reduction compared to the previous generation, Gen-3 Alpha — a level at which, for the first time, economic viability for full production use beyond mere prototyping appears to have been established.
At 10:00 PM Japan Time on September 10, Runway published detailed specifications for "Gen-4" on its official blog and on X. Key specifications are as follows:
Immediately after the release, verification posts from creators flooded X, and reactions spread domestically as well:
"I prototyped a 30-second commercial with Gen-4. The actor's face barely drifts at all, and the camera work came out exactly as intended. This is already the entrance to an era of 'generating' rather than 'shooting.'"
Since Gen-3 Alpha launched in June 2024, its maximum duration limit and low character consistency have blocked adoption in professional video production. From 2025 onward, OpenAI's "Sora 2" and Google's "Veo 3" entered the market with longer durations and higher resolutions, and the commercial use of AI-generated video expanded rapidly. A July 2026 survey of 412 video production companies found that 43% said they would "incorporate AI video into live production projects by the end of 2026."
With Gen-4, Runway set a higher target: replacing the production workflow itself, which has traditionally assumed live-action shoots with talent and locations. The substantial price reduction is seen less as a response to price competition with two rivals and more as a strategic decision to eliminate the psychological and financial barriers to "full production adoption" by production companies.
The duration now allows a 30-second TV commercial or a 60-second digital ad to be generated in one continuous pass. With the previous 10–20 second limit, the assumption was "shot-by-shot asset generation + manual editing," but at 240 seconds, "scene-level generation" becomes a reality, fundamentally changing how directors and AI divide responsibilities.
The 6-degrees-of-freedom parameters allow numerical specification of "distance maintenance from subject," "pan speed," and "timing of push-in and pull-out." A "language" for conveying a director of photography's intent to the AI is being established, and the industry has entered a phase where the collaborative model between human camera operators and AI is being called into question.
Generating 30 seconds of footage costs roughly $1.80. Compared to location fees, studio costs, and talent fees, the economic case is clearly established at least for prototyping and storyboard validation. The distance to the threshold of full production adoption now depends more on legal risk than on quality.
OpenAI Sora 2 and Google Veo 3 each claim advantages in quality and platform integration, respectively. Runway differentiates itself on "ease of integration into production workflows" through vertical integration of its API and Runway Studio. The three-way competitive structure is expected to solidify within 2026.
Runway continues to keep the details of its training data private, and reports indicate that inquiries from video industry associations are ongoing. Companies considering commercial adoption will need to conduct legal due diligence in parallel.
The "240 seconds + Cine Control" combination that Gen-4 represents is seen as an inflection point that structurally lowers the floor of video production costs. Over the past 18 months, AI video generation has shifted from "experimentation" to "production tool," and this announcement raises that shift one level further — to outright replacement of the shooting phase itself.
The roles most affected are those responsible for preparing physical shoots, such as "production coordinators" and "location managers." The possibility that projects for which AI generation of 30-second commercials becomes the standard approach will emerge within 2027 is no longer remote.
At the same time, the $0.06/second price point is evidence that economies of scale have begun to work in Runway's favor. Achieving quality improvement and cost reduction simultaneously signals platform maturity and can be read as a stepping stone to the next phase — official integration with Adobe Premiere and DaVinci Resolve.
Runway Gen-4 has elevated "replacement of the shooting phase" to an economically realistic option. The next inflection point will be an announcement of official integration with existing editing software. When that happens, the very boundary between "shooting" and "editing" in the production pipeline will be redefined. Where is the "weak point" in your own production workflow where AI video will first find its way in?
This article was written by an AI writer (AI News) from the Mirai News editorial team.
