Text-to-video conversion technology developed by companies like Runway and nVidia could wreak havoc on Hollywood and spoil our senses. is real?
In the last few years, AI models that can call up original images from a few lines of text have taken the internet by storm, flooding our online lives with everything from elaborate fantasy landscapes to eerie landscapes. . digital devil, from bizarre alternate history to photojournalism. At the same time, the rise of another digital technology, deepfakes, has allowed celebrities’ faces to be glued onto bodies that aren’t their own and inserted into anything. porn To Dystopian disinformation campaignBoth technologies have played a role in transforming creativity and ideas about creativity. undermine the credibility of visual mediabut the turmoil does not appear to be slowing down anytime soon.
The race to develop AI generators capable of outputting not only still images but also realistic videos has been going on for some time. In December 2022, we got an early preview of what it might look like thanks to the surreal sitcom nothing, forever, generated sketchy pixel art graphics with the help of several AI systems, including the large language model GPT-3. However, in the last few weeks, with the launch of the first real text-to-video models, the competition is really heating up.
Basically, the text-to-video work is kinda scary visuals Will Smith voraciously eats a plate of spaghetti in early April. It’s not just his videos that go viral, a product of his ModelScope, a text-to-video system developed by Hugging Face’s collaborative team. See also dwayne johnson munching on rocks, Arnold Schwarzenegger punch pizza, giggling brother Playing a guitar solo in front of an erupting volcano, it happened to be an example that captured the public imagination, making it the ugly face of the coming AI video revolution. But just a month later, its face looks more real, and a revolution looks more and more imminent.
With several companies competing to produce the most realistic AI videos on the market, there are some obvious concerns. What does this mean for the already skyrocketing rate of misinformation online? What would the dramatic landscape look like? Is Hollywood doomed? Below you will find some answers.
By now, you’ve probably heard (or tried) some of the biggest text-to-image generators, including DALL-E 2, Imagen, and Midjourney. If so, the basic concept is familiar: you enter a few lines of text prompts describing an object or situation, the style in which it was drawn, as well as details such as angles and camera lenses, and the generator Watch it come to life. in seconds.
From the user’s point of view, the text-to-video generators are essentially the same. As a “creative suite” that utilizes AI Runway said: If you could say it, now you can see it.
Text-to-video conversion has been around for some time as a concept. In September 2022, Meta will launch an imaginatively titled make a videofollowed shortly by Google’s announcement Image VideoThe problem is, all they had to share was a research paper and some pre-made examples.
Like the previous text-to-image model, text-to-video generators are designed to help regular people (i.e. not just AI experts) understand how to use them to create interesting things. It’s almost exploded now. Or strangely beautiful, or realistic enough to be funny. In this way, complex creatives tend to spread his technology into the mainstream. Obvious ethical issues tend to follow soon after.
Certainly no.they are definitely not equal sand dunesHonestly, the best AI-generated video couldn’t even hold a candle to my least favorite Marvel movie. Remember what the AI-generated images looked like when they first arrived? fuzzy blob and Rugged, ill-shaped imitations The work of a real artist? Now, just a few years later, they are mistaken for real photographs left, right, and center, and are used to push the boundaries of human creativity. AI is starting to understand the notoriously nasty tricks.
Like it or not, there’s no reason to expect the upward curve of text-to-video improvements to be similarly less steep. Especially as more and more competitors enter what is likely to be a highly profitable race to produce commercially viable results.
Runway, one of the two tech startups behind the controversial AI art generator Stable Diffusion, has begun its public testing. 2nd generation Last month’s video model showed amazing results across social media. Gen-2 users can create videos from scratch by entering simple text prompts. You can also offer the option to incorporate prompt images such as portraits (think: deepfakes, but with full creative control).
ModelScope, on the other hand, is a relatively basic model created by the research arm of e-commerce giant Alibaba. The company, which created the bizarre, short-form clips that marked the early days of the text-to-video topic, has also been controversial for the fact that many of the videos produced include blatant shutterstock watermarks. image scraping sauce.
And, of course, there are bigger players like Meta and Google. But despite several eerie photorealistic Mark Zuckerberg avatars circulating on the internet lately, both have remained relatively quiet since their first research papers were published last year. Another promising research project (or sinister one, depending on your outlook) is coming courtesy of software company nVidia. promised We plan to reveal more at our meeting in August.
Others combine multiple AI tools to generate more compelling AI videos. For example, generate images in Midjourney and bring them to life via Runway. Also, this high-resolution (albeit less animated) beauty, reconsideration of Harry potter As a Balenciaga fantasy, or fake trailer Wes Anderson remake Star Wars.
At this point, it’s almost undeniable that text-to-video generators are coming, and like image generators before them, they will have a huge impact on the media landscape. ) to confusion about the veracity of what is in front of them, expect more of basically the same reaction.
Of course, text-to-video proponents would argue that there are many advantages as well. From cameras to digital film to CGI, new technologies are constantly expanding our artistic horizons, for better or worse, and AI video has the potential to unlock entirely new forms of expression. If we can’t do that, we might at least be able to customize our own TV shows to keep us hooked even after our jobs are no longer automated. , we might even be able to sneak preview our demise at the hands of robotic overlords.
