The Film Sets of the Near Future: As Artificial as the Film Itself

To grasp the seismic impact generative AI is poised to have on the film industry, this brief video says it all. Just one shot is truly “real“—and that includes the so-called “behind-the-scenes footage.”

If you didn’t look too closely, this playful production could easily be mistaken for a genuine advertisement—one that appears to have a remarkably high budget, judging by the impressive locations featured.

An AI-Generated Vision

Of course, it’s not a real ad; the vision is almost entirely AI-generated, the voiceover comes from a Fiverr freelancer, the music is sourced from a stock library, and the entire production—including the bonus behind-the-scenes footage—was created by a single producer/editor in roughly three weeks using Google DeepMind’s Veo 2 text-to-video system.

In fact, the only genuinely “real” shot in the entire piece is the talking head in the behind-the-scenes segment, featuring creator László Gaál, who appears to be based in Vietnam and orchestrated the entire project. Take a look:

Gaál’s Experience with Veo 2

Gaál has early access to the Veo 2 system, which is not yet publicly available. In a Reddit thread, he explains that generating the ad took about 12 days, with an additional four days dedicated to creating the behind-the-scenes footage.

The main challenge, he notes, is maintaining consistency in scenes and characters, as Veo 2 currently lacks the ability to use reference footage or images: “Right now, you can only use text, so you need to ensure the characters remain consistent… to convince the viewer [it’s the same person].”

The use of short clips in the project is a stylistic choice rather than an attempt to mask any limitations of the AI’s longer sequences. “It’s more about serving the story [than hiding poor AI generation],” Gaál writes. “The full clips look good too—I might share some on Twitter later.”

Veo’s Game-Changing Potential

From a production standpoint, this is astonishing. Gaál used Veo to generate footage that would have required millions of dollars and a sizeable team to produce in the real world—highlighting that now might not be the best time to start a career in the film industry.

From a viewer’s perspective, once character consistency is resolved, the primary distinction between AI-generated scenes and “real” Hollywood productions might be that the AI-generated ones actually look better. To illustrate, here’s another test clip from Gaál, shared five days ago:

Impressive work. Here’s the kicker: Veo 2 is still at a stage comparable to ChatGPT, handling only small portions of a project. Looking ahead, the trajectory of this technology is unmistakable.

In the not-too-distant future, video generation AI will likely handle entire projects autonomously—writing detailed movie scripts using language models, generating video, audio, voice, and music assets, and then assembling and editing everything. Looking even further ahead, it could create fully edited, scored films with sound effects and voices in one seamless process. At that point, anyone could make their own movie.

The Dawn of Personalized, On-Demand Entertainment

Once this becomes fast, affordable, and user-friendly, we’ll enter an era of personalized entertainment. You could request tailored content about anything you imagine and adjust every detail almost in real time. For example: “Hey Google, create a show about a six-year-old who won’t clean her room, gets lost under a pile of laundry, and has to team up with mice and spiders to plan an escape.”

From there, when a system like Veo integrates with something like Google Genie, the experience could become fully interactive, real-time, and 3D, ready for use with VR goggles. At that stage, we’re essentially looking at an on-demand Star Trek-style holodeck.

That’s the utopian vision. On the dystopian side, it means the end of trusting what we see with our own eyes. The quality of video these systems can now produce has leapt from laughable to breathtaking in just a few months.

Google and other major AI labs will likely implement digital watermarks to label their files as AI-generated. However, the past year has demonstrated that nearly every groundbreaking AI innovation is swiftly replicated by numerous other organizations—some of which may not be as trustworthy.

As a result, it’s becoming increasingly impossible to trust any video as authentic.

The Impact of Unlimited On-Demand Entertainment

Looking at the situation from an even more dystopian perspective, consider how divided society already feels today, in an era where people have nearly unlimited access to on-demand entertainment, rather than everyone sharing the same TV show experience the night before school.

Now, imagine how much more fragmented things could become if everyone starts custom-creating their own entertainment, becoming even more entrenched in their personal beliefs and ideologies.

Then, consider one of the key ideas from Yuval Noah Harari’s insightful book Sapiens: that “shared fictions” such as money, nation-states, religion, and law are humanity’s most powerful tools for control and social order. You might also have read his latest book Nexus, which argues that by taking over our writing, communication, and soon our entertainment, AIs have essentially “hacked the operating system of our civilization,” with potentially catastrophic consequences for humanity.

Or… you might simply enjoy the stunning visuals and think about the kinds of movies you’d create if all barriers to creativity were removed.

Read the original article on: New Atlas

Read more: Images of Artificial Intelligence Tend to Exaggerate Stereotypes