AI video startup Runway has just released its much-anticipated Gen-4 video synthesis model, claiming to finally tackle the longstanding challenges in AI video generation. This groundbreaking update promises to deliver greater consistency in characters and objects across shots—an issue that has plagued AI-generated films since their inception.

If you've had the chance to witness AI-created short films, you'll understand that they often resemble surreal dreamscapes more than coherent narratives. It's not uncommon for characters or objects to change appearance between scenes, disrupting any sense of continuity. However, Runway assures users that Gen-4 can maintain consistent representations, provided filmmakers upload a reference image of the character or object they wish to use in their projects.

The company has showcased example videos featuring the same individual appearing in diverse scenes, as well as consistent representations of objects, regardless of changing environments and lighting conditions. This advancement means that filmmakers can now achieve seamless coverage of the same environment or subjects from various angles within a single sequence—something that was virtually impossible with previous models, Gen-2 and Gen-3.

Gen-3, launched just under a year ago, had offered improvements by extending the video length from two seconds to a more usable ten seconds, although it struggled with maintaining visual coherence across different shots. With Gen-4's enhancements, users are optimistic about a significant leap forward in the quality of AI-generated content.

Runway carved a unique niche for itself amidst fierce competition in the AI video landscape. Founded in 2018 by art students from New York University's Tisch School of the Arts, the company was one of the first to introduce a usable video-generation tool for the public. While better-funded competitors like OpenAI lead the market, Runway has chosen to cater specifically to creative professionals—designers, filmmakers, and content creators—by positioning itself as a supportive tool within existing creative processes.

This strategy has paid off, exemplified by a partnership with motion picture giant Lionsgate, allowing the startup to lawfully train its models on Lionsgate's film library while providing tailored tools for production and post-production. However, this approach has not been without criticism, as Runway, along with other companies like Midjourney, faces litigation from artists alleging that their work was misappropriated for training AI models.

While the specifics of the data used to train Gen-4 remain largely undisclosed, a report from 404 Media hinted that the training set might include videos scraped from popular YouTube channels and film studios, raising further concerns about intellectual property in the AI industry.

As for the rollout of Gen-4, while it's now available to all paid plans and enterprise customers, some users have reported delays in access. Prices for individual plans start at $15 per month, can reach up to $95 per month for the most extensive features, or $1,500 per year for enterprise accounts. Interestingly, there’s also an “Explore Mode” in the highest tier plan, allowing users to experiment with unlimited generations, helping them refine their creative output before making a financial commitment.

In a landscape crowded with competition, all eyes are now on Runway and its Gen-4 model. Will it live up to the hype and fulfill the demands of creative professionals, or will the hurdles of AI video generation continue to prove insurmountable? As we await widespread user access, one thing is certain: the future of AI in filmmaking is evolving rapidly, and Gen-4 could be a pivotal moment in that journey!