AI Video Generation: Help & Competition for Creators

AI generated images

NOS News

  • Olivier de Neve

    Editor Economy

  • Olivier de Neve

    Editor Economy

A race on huge bath ducks in a full stadium or a cowboy that stands with legs spread on two galloping horses. Clearly fantasy scenarios, made with video software that runs on artificial intelligence.

The boundary between fake and really is becoming increasingly vague. Moreover, the software can also be used for videos that could really have been.

This week, OpenAi, the company behind Chatbot Chatgpt, presented the second version of video generator Sora. OpenAi promises that this version makes even more realistic videos than its predecessor. And where the first Sora only delivered images, version 2 also offers videos with sound, such as voices, music and sound effects. The software can now only be used by Americans and Canadians with an invitation. The rest of the world will follow later.

Sora 2 is the newest step in a race between tech giants for the best AI video software. OpenAi’s big competitor is Google, which in May of this year with VEO 3 again achieved a lead over Sora. Elon Musks company Xai joined the battle in August, with Grok Imagine.

Radical

Video makers of flesh and blood, such as employees of advertising companies, feel the pressure. AI tools promise to do the work faster and cheaper than human makers.

Advertising agencies are now usually paid per hour, but that is under pressure by AI. “The hour-invoice approach is creaking, while AI and tech companies are taking over more and more tasks,” warned industry association through last month.

The need for renewal is clear for advertising agency Monks, director Vanessa Hofland-Noordervliet explains. “The videos of AI tools now see mindblowing Fantastic out. Our expectation is that that development will continue. That is why we have been radically changing our business model for a while. For example, by looking at a subscription model instead of working per hour. We then build the AI ​​tools that our customer can use, designed for the brand or company. “

Monks also checks which AI aids there are every week, and which are best workable. “Now that is still VEO 3 for making videos, but that could be Sora 2 next week,” says Hofland-Noordervliet. “We have decided to put the fear of AI next to us and embrace the tension. But it will remain super scary of course.”

Luke Andries agrees. With his company Aigenwijs, he gives training in the use of AI to the business community and government. “Many advertising agencies look at this with nodding knees. If you have a small desk or deliver mediocre videos, you can close. Companies can just make their videos themselves.”

Imperfections

With slick examples, Sora, Veo and Grok show what is possible with their tools. But how good are those AI videos now? “Gravity and physical movements were always difficult,” says Andries, who could already use Sora 2. “But that too is now becoming bizarre realistic.”

However, the videos of Sora 2 cannot take longer than 10 seconds, Andries explains. “That has to do with, among other things, the amount of computing power that is needed and the corresponding costs. For longer AI-Videos, it is consistent to portray characters or environments. Objects are suddenly in a different place. The software does not always do what you want. But sometimes you can use it up with some creative tricks.”

Despite such imperfections, the risk of disinformation and propaganda by AI videos is increasing. “Very problematic,” says Olya Kudina, assistant professor ethics and philosophy of technology at TU Delft. “In addition, it also has a negative influence on the creative industry. Advertising is by definition a form of manipulation, for example to influence people to buy a product. People are used to traditional advertisements. But if you also do artificially made images and sound, you no longer know what real is.”

Kudina therefore argues for clearer labels in, for example, AI advertisements. “Simply prior to a video saying that it was made with AI, is not enough. You need a continuous memory in the picture.”

Thanks to the video generators, everyone can make good looking videos, emphasizes Monks director Hofland-Noordervliet. “My neighbor can therefore also be creative at the level. So no home-garden-and-kitchen movies, but really at feature film level.”

Related Posts

Leave a Comment