The AI video generation race has recently been heating up, with major tech players and emerging startups all eager to create models that can seamlessly produce high-quality, realistic videos from simple text prompts. After OpenAI's flashy release of its video tool Sora, industry giants like Google and up-and-comers like Runway couldn't sit idly by. They quickly rolled up their sleeves, intent on developing their own models that could rival this new wave of GenAI-powered video tools.
On the surface, these AI video tools appear to offer a "cost-effective and time-saving" revolution for filmmakers, but beneath the shiny exterior, there are a few potential landmines. First, the issue of copyright is starting to rear its ugly head, and it could soon trigger a major legal battle. Second, the risk of deepfake videos flooding the internet is real, and the chaos that could ensue is enough to make anyone nervous. Additionally, as competition in the AI sector becomes more intense, some companies seem to be losing focus, rushing to meet deadlines while sidelining security concerns. To make matters worse, AI video generation consumes a significant amount of energy, far more than text generation, putting a considerable strain on power grids.
Google has clearly refused to be left behind. Just a week after the grand release of Sora, the company unveiled its own GenAI video tool, Veo 2. However, Veo 2 is a bit exclusive—it's only available to early access users in the U.S. who are 18 or older, and Google has kept the details on how it works and its pricing close to the chest. While Veo 2 can generate video clips of up to two minutes, don’t get too excited—it's not without its flaws. The video consistency can be poor, often resembling a patchwork of mismatched images, and handling complex actions can be a struggle. Despite these shortcomings, many still view Veo 2 as a "promising contender" and one of the most exciting AI video tools in the works.
Meanwhile, OpenAI had quietly rolled out the beta version of Sora to a select group of testers back in February 2024, and by December, the tool was made available to all ChatGPT Plus and Pro subscribers. Sora comes bundled with the subscription, offering access to text, code, image generation, and video creation all in one package—a "one-stop shop" for those who want it all. Sora’s focus is primarily on short videos aimed at social media and marketing, allowing users to generate a 20-second clip by simply typing a description. However, it's not without its flaws. The tool often trips up when it comes to basic physical principles, leading to errors that range from amusing to frustrating. Even YouTube's Marques Brownlee couldn’t resist pointing out that Sora generated a fake plant in his video that was nearly identical to one in his original footage—a potential copyright headache.
Looking at the startup scene, Runway has made a strong mark with its Gen-3 Alpha model, which became available in all paid plans starting in September 2023. Runway’s tool was instrumental in the creation of the Oscar-winning film *Everything Everywhere All at Once, contributing heavily to its special effects. The subscription plans range from $144 to $1,500 per year, depending on how much video content the user wants to generate, providing options for every budget. While Runway was the first company to release such a tool publicly, early market entry doesn't always guarantee success, and being first doesn't mean you're safe. Thankfully, Runway has been proactive, forging partnerships with creators early on to prevent copyright issues. In December 2023, it teamed up with Getty Images to offer a more "commercially secure" version of its tool. By September 2024, Runway joined forces with Lionsgate Studios to enhance its AI training, and in December, it launched an independent creator network, giving creators a platform to showcase their work to production companies and studios. Through these strategic moves, Runway has secured its place in the AI video generation space, establishing a solid foundation for future growth.
Ranking
