AI video generation is rapidly evolving, as evidenced by the prevalence of viral videos made with algorithms in recent weeks. Video memes featuring Harry Potter reimagined as a Balenciaga commercial and surreal footage of Will Smith eating spaghetti have captivated viewers worldwide and show how quickly technology is advancing. Similarly to AI-created images, and the program Craiyon (formerly known as DALL-E Mini) which let anyone generate recognizable images, AI video generators are also becoming increasingly accessible and sophisticated. After making headlines when AI-made images purporting to show Donald Trump’s arrest went viral last month, Midjourney temporarily stopped offering their service for free.
Runway ML, a startup that is developing AI tools for professional image and video creation and editing, recently launched a new technique for more efficient video editing. They offered an example of a dreamlike footage of a cat walking through a ‘cloudscape’ created with existing footage in just minutes. Another company, Luma AI, is using a technique known as neural radiance fields to turn 2D photographs into detailed 3D scenes.
AI video creation has reached a crucial tipping point and the technology’s advances are allowing us to create and edit videos more efficiently than ever before. A growing wave of artificially-created videos has the potential to blur the boundaries between real footage and fake or to give rise to new visual tools. Nonetheless, creating convincing videos with AI still involves computational strain and their quality is yet to compare to the level achieved with images. This reveals the need to consider the implications of AI video-making, particularly its capacity to propagate ugly stereotypes and to be used as a propaganda tool.
OpenAI is the organization behind ChatGPT, an AI image generator that ranks among the earliest to show that AI can take a text prompt and turn it into what looks like real pictures and illustrations. Although it was when carefully restricted, after Craiyon was made open source, DALL-E also became freely available and programs from Midjourney and Dream Studio have been subsequently developing and refining image-making tools for ease of use and higher quality. Popular director Darren Aronofsky is a fan of Runway ML, which was used by the visual effects team at Everything Everywhere All At Once, the Oscar winning production.
It is clear that AI video generators are nearing a milestone as they become easier to access and their potential increases. We must be aware of their practical applications as well as potential risks, such as their malleability to propagate fake news or to resurrect stereotypes. Though current AI-made videos have their momentary glitches, the technology is rapidly evolving and its far-reaching effects may only be a matter of time.
