Hand-on-heart what did you think of the video above? You can be honest – since it was largely produced by an algorithm, there’s zero chance of the director offering to beat you up.
The video was created using an online text-to-video platform called VideoGen. It cost about the same as ordering a pizza and the whole process, including the time spent writing the script and tinkering with the output, probably took around 45 minutes. The time it took an AI algorithm to generate the video itself was under a minute.
For the most part, it’s perfectly serviceable, no? Especially when you consider how time-consuming and expensive it can be to create video content from scratch. (The average cost of even simple content, like an explainer video, is around $5,400.) Unless you were looking for it, I think you’d be hard pushed to distinguish this from most other corporate marketing videos that combine stock imagery, ambient music, and voice over.
So, as 18% of businesses weave AI tools into their video production workflows, is AI-generated video content worth considering?
Understanding AI tools for video creation
AI is already transforming video production in the same way it’s transforming other complex processes, by automating and accelerating repetitive and time-consuming tasks. But we wanted to go a step further and look at the tools marketing teams without their own video production capabilities (like our own) can use to create full videos from scratch.
There’s a lot of variety in this market, in terms of both video style and content quality.
Some programs use prompts to generate “original” footage (welcome to the uncanny valley, please mind the hands). Others provide digital avatars that will read your script like some sort of slightly unsettling Red Dwarf throwback.
We created our video with VideoGen, because we wanted a video that wasn’t so obviously AI-generated. (And, let’s be frank, because it was quick and cheap.) And because its own marketing hails it as “the most powerful AI video generator ever!” All VideoGen asks is that you enter a script into a text box. The AI then uses key words from your text to pull images from a stock footage library, while a computer-generated voiceover reads it aloud.
The process is more “sortamated” than automated, as it rarely gets everything right first time. As you can see from the first version of our introductory video below, you’re often met with irrelevant imagery, inappropriate music, mistimed transitions and a host of other not-quites. At one point the program insisted on putting a “like and subscribe” page for YouTube directly in the middle of the video, which felt hopeful at best – like AI’s version of *shrug* “maybe this’ll do”.
All of this means, even with one of the most basic tools around, marketers are destined to spend a fair amount of time using a software interface to fine tune their creations.
The dangers of undifferentiation
As I put various iterations of our script through the program, I began to get an odd sense of déjà vu. It quite quickly became apparent that I was being fed the same stock footage over and over – same clip, different day. Despite the service boasting a library of 500,000 videos to choose from.
This seemed far from ideal when the last thing you want your marketing to be is generic and derivative.
And that’s a problem that extends beyond stock-footage-dependent tools like VideoGen. Even if you use a tool that generates images from scratch, you’re likely to encounter a “look” that can be attributed to that particular program. Much the same way lots of people can instinctively tell when something has been written by Chat-GPT.
For me, this is a good enough reason never to use AI-generated video for customer-facing marketing content. Above all else, you want that content to be both unique and specific to your brand.
But there are some scenarios where it could save marketers time and money, with far fewer risks. Given many people find video content much easier to engage with than the written word, AI-generated videos could be a cheap and effective addition to the usual package of sales enablement collateral.
Equally, by providing a way to create rough prototypes of video content, they’re likely to help marketers better convey their ideas when briefing in their human video production teams.
Cutting edge or cutting corners?
Marketers have spent decades talking about the importance of personalised journeys and tailored content. And in recent years, the rise in data analytics has made engaging with prospects as people, not percentages, much more achievable.
In many ways, presenting your audience with AI-generated content runs counter to this trend. Yes, you could argue it creates the opportunity to cost-effectively spin up multiple videos, within a small window of time, and speak more directly to specific audience segments – or even specific individuals. But if they can tell it’s been made by AI – and they can, even when it’s been done well – how will that affect their perception of your brand? Toys “R” Us recently provided the world with a great example of how AI-generated content can backfire.
This won’t be true in every case. In some instances and industries, creating video content that shows you’re operating at the cutting edge of AI experimentation could have the opposite effect, impressing your customers and prospects and positioning you as a trailblazer. But whatever your brand and your target audience, the decision to use AI-generated video isn’t one to be taken lightly.
What you won’t get from AI-generated videos
As marketers, we strive to create moments of resonance – genuine, human reactions. Most often, this happens when we break with the long-established formula. When, instead of slotting the same old ideas together in the same old ways as everyone else, we find more surprising (and usually, more human) messages and media.
Those moments require, perhaps more than anything, a sophisticated understanding of human relationships and emotions. Something that can’t be quantified or taught, but is instead just inherently understood – by people, but not by algorithms or machines.
Talking this blog post through with my colleagues, I kept finding myself coming back to the ending of Mike Nichols’ The Graduate (1968). In the final scene of the film, Dustin Hoffman and Katherine Ross’s characters sit at the back of a bus having fled from Ross’s wedding.
There’s a pivotal moment here – what in screenwriting parlance is called a beat – that’s dragged out over several heavy seconds, as we slowly see the enormity of our characters’ actions dawn on them. You’ve probably seen it, so I won’t go on. But this moment, for me, is the perfect example of a human artist’s ability to connect with an audience by understanding the human condition – and being led by what felt right.
Nichols intuitively knew how long that wordless moment needed to be. He understood what his characters were thinking in those seconds, and how the audience would understand those thoughts, too. If you asked AI to edit that film, it would almost certainly be six seconds shorter, and 100 times worse. It would be technically correct. But at the same time, entirely wrong. It’s in this intangible space – in those insightful, empathetic six seconds – that humans will continue to show their creative superiority.
In time, I have no doubt that a healthy balance will be found between marketer and machine – one in which AI simplifies and accelerates processes and leaves more room for such human creativity to flourish. We’re already seeing the fruits of that fertile middle ground. But relying on AI to be creative on your behalf will always feel like cutting corners to me. And when its source material is solely things people have created in the past, you’re removing any hope for originality.
For now then, we should continue to put people in charge of talking to people. And only allow technology to lend a hand.
