Runway Gen-4 solves the biggest problem in the artificial intelligence video: character consistency across the viewer

Runway Ai Inc. She launched the most advanced video generation model today, as she entered the next stage of competition to create tools that can transform movies. New Gen-4 system The consistency of the character and the scene offers multiple shots – the ability that escapes from most AI’s video generators so far.

New York -based emerging, with the support of this Googleand NafidiaAnd Salesforce“Gen-4“For all paid subscribers and institutions agents, with additional planned features later this week. Users can create five and ten seconds with 720 pixels.

The version comes after only days Opeenai images A created feature Cultural phenomenonWith the request of millions of users Studio GHibli-Style Pictures through ChatGPT. The viral trend became very common, as Openai’s servers were temporarily crashed, as CEO SAM Altman tweeted that. “Our graphics processing units melted“Because of the unprecedented demand, Ghibli’s Gibli images also raised hot discussions on copyright, with a lot of questioning whether artificial intelligence companies can simulate legally distinctive technical methods.

https://www.youtube.com/watch?

The continuity of vision: the missing piece in the film industry so far

The consistency of the character and the scene – maintaining the same visible elements through multiple shots and angles – was Achilles heel to generate artificial intelligence video. When the character’s face changes skillfully between the discounts or the background element, it disappears without explanation, the artificial nature of the content becomes immediately clear to viewers.

The challenge stems from how these models work on a basic level. The former artificial intelligence generators dealt with each frame as a separate creative task, with only loose connections between them. Imagine to ask a room full of artists for each one frame of a movie without seeing what happened before or after – the result will be visually disassembled.

Runway’s Gen-4 It seems that you may deal with this problem by creating up to a continuous memory of visual elements. Once created a letter, object or environment, the system can be presented from different angles while maintaining its basic features. This is not just an artistic improvement; It is the difference between creating interesting visual excerpts and actual stories.

Using visual references, along with guidelines, Gen-4 allows you to create new photos and videos with consistent patterns, topics, locations and more. Allow continuity and control your stories.

To test the form of the form of the model, we assemble … pic.twitter.com/iyz2Baew2U

Runway (runwayml) March 31, 2025

According to Runway documents, Gen-4 allows users to provide reference images of topics and describe the composition they want, with artificial intelligence generated consistent outputs from different angles. The company claims that the model can present videos with realistic movement while maintaining the consistency of the topic, subject and elegance.

To display the capabilities of the model, Runway released many short films that were fully created with the Gen-4. One movie, “New York is a zoo“Explains the visual effects of the model by placing realistic animals in the New York Film Settings. Other entitled” Bat “Recovery“Explorer who is looking for a mysterious flower and produced in less than a week.

https://www.youtube.com/watch?

From the animation of the face to the global models: the development of the film industry from Ai Runway

Gen-4 depends on the previous tools of the runway. In October, the company issued One workIt is an feature that allows film makers to capture facial expressions from the smartphone video and transfer them to characters created from artificial intelligence. The following month, Runway added an applicant 3D camera controls For the Gen-3 ALPHA Turbo model, allowing users to enlarge and get out of the viewer while maintaining the letters of letters.

This path reveals the strategic vision of the runway. While the competitors focus on creating one more realistic images or clips, Runway has been assembled by the components of the full digital production pipeline. This approach looks closer to how actual directors work – tackling performance, coverage and visual continuity problems as interconnected challenges instead of isolated technical obstacles.

The development of face -to -face animation tools indicates that Runway realizes that the AI ​​-backed film industry needs to follow the logic of traditional production to be really useful. It is the difference between creating an experimental offer for technology and construction tools that professionals can actually integrate into their workflow.

Amnesty International Video Battle of Millions of dollars is rising

Financial effects are great for the runway, which norms the reports New financing round This would estimate the company at $ 4 billion. According to financial reports, the startup aims to reach $ 300 million in annual revenue This year, after the launch of new products and an application programming interface for their video generation models.

The runway is followed by Hollywood partnerships, Securing a deal with Lionsgate To create a dedicated AI video generation model that depends on a studio catalog of more than 20,000 titles. The company has also established One hundred film boxesFilm makers offer up to one million dollars to produce movies using artificial intelligence.

“We believe that the best stories have not been listed yet, but traditional financing mechanisms often overlook new and emerging visions within the largest ecosystem in the industry,” explains Runway on its website.

However, technology raises fears for the film industry. A 2024 studies Employed from Animation Syndicate I found that 75 % of the film production companies that adopted Amnesty International have reduced, unified or canceled jobs. Study projects that more than 100,000 entertainment jobs in the United States will be affected by the Instelligence IQ by 2026.

Copyright questions follow the creative explosion of Amnesty International

Like other artificial intelligence companies, Runway faces legal audit on their training data. The company is currently defending itself in a lawsuit filed by artists who claim that their copyright work has been used to train artificial intelligence models without permission. Runway cited the doctrine of fair use as a defense of it, although the courts have not yet ruled this application of the copyright law.

Publishing debate intensified last week with Openai Ghibli Studio featureWhich allowed users to create pictures in a distinctive style of the animation studio in Hayao Miyazaki without explicit permission. Unlike Openai, who refuses to generate photos in the style of artists but allows studio styles, Runway has publicly not explained his policies on imitating elegance.

This discrimination is increasingly arbitrary as artificial intelligence models become more advanced. The separation line between learning from the wide technical traditions and the copying of the specified creative patterns has been clarified to the near. When AI can simulate the visual language that took decades from MIYAZAKI to develop it, does it matter whether we were asking to copy the studio or the artist himself?

Upon interrogation about training data sources, Runway refused to provide details, noting competitive concerns. This ostrich has become a standard practice among the developers of artificial intelligence, but it is still a point of disagreement for the creators.

While exploring marketing agencies, educational content creators, and corporate communications teams Gen-4 Video production can be simplified, the question is transmitted from technical capabilities to creative application.

For directors, technology represents the opportunity and disruption. Independent creators can access the possibilities of visual effects previously available only for the main studios, while traditional VFX professionals and the future are not sure.

The uncomfortable truth is that technical restrictions have never prevented most people from making convincing films. The ability to maintain the continuity of vision will suddenly create a generation of geniuses of telling stories. However, what might do is to remove adequate friction from the process that more people can experience visual narration without the need for specialized training or expensive equipment.

Perhaps the most deep aspect of Gen-4 It is not what it can create, but what it suggests about our relationship with the visible media to move forward. We enter a era in which the bottle neck does not have a skill or technical budget, but imagination and purpose. In a world anyone can create any image that he can describe, the important question becomes: What is worth displaying?

When we enter a afternoon as it requires creating a movie more than one reference and demanding, the most urgent question is not whether AI can create convincing videos, but whether we can find something meaningful to say when there are tools that say anything within our hands.