Calling it “text-to-video” undersells what’s actually going on in the latest innovation from OpenAI.
The company’s Sora programme allows for the creation of “realistic and imaginative scenes” from simple text inputs.
An example shown on the company’s website is a 25 second crane shot that pushes through a sepia-toned 19th century American town with dirt roads, grazing horses and a variety of wooden buildings.
The first seconds are particularly striking as we move through a creekbed, flying towards low hills dotted with scrubby pine. There’s so much detail, and it’s hard to believe the only given prompt was “Historical footage of California during the gold rush”.
Unveiled last week but not yet made available to the public, the very existence of Sora has caused waves of discussion and analysis. American director Tyler Perry said he was pausing work on an $800 million expansion of his studio in Atlanta, primarily because of the increasing availability of AI tools.
As with text-based generative AI last year, we in financial marketing and communications must reckon with what this latest technology means for our practice. I’m broadly excited for the possibilities, although I also want to sound a few notes of caution.
Video production is complex, requiring lighting, sound and cinematography tools for live action and design power and programmes for animated videos. The creation of a minute long video can easily take months and cost tens of thousands of pounds.
The potential to revolutionize video production is huge, by leveraging creative expression, providing high-quality videos and making bespoke video production accessible to all.
This doesn’t mean a total flattening of the creative value chain, however. Even in the video of that California town, things in the second half start to seem less lifelike and more Playstation 4 game when we get closer on the town. Fully AI videos won’t have the ability to create fully professional environments, at least not in the short-term.
What it can do is help with the development process. Storyboards only show snapshots of the action, and it can be hard for our clients to really imagine the final projects.
Sora and similar tools will enhance our creative team's productivity by accelerating video production and generating quick demo ideas that we can use to develop our main video projects.
Similarly we can create a piece in a presentation that can help developing pitches and visualising ideas for campaigns. One of our biggest challenges is moving a concept from the brain of a creative into a physical medium where others can understand our vision. Easy to generate, highly customizable generated video can help there, allowing us to increase audience engagement and attract new prospects.
We may be able to move away from relying on overused stock videos for certain establishing shots. Many productions don’t have the budget to, for example, show a view of the New York City skyline from the top of Times Square. This means that commercially licensed footage is shown over and over, lessening its impact. Now we may be able to generate brand new pieces and new angles.
We are going to enter a period of rapid experimentation as people see where these tools can actually add value. While we encourage playing around, stay away from these pitfalls.
First, don’t use video tools to create deepfakes. Trying to create a copy of someone’s likeness is increasingly frowned upon. You risk being banned from tools – and increasingly open yourself upon to civil and criminal charges. OpenAI recognises the potential dangers and is deploying “red teamers” – red teamers — domain experts in areas like misinformation, hateful content, and bias – to test the platform.
Remember that generative AI is ultimately a tool to allow you to showcase your own creativity. You cannot expect Sora to do the hard work of coming up with something truly new. If it can be created solely by the algorithm, it cannot be truly tailored to your client and end customer.
But in sum there’s plenty to be excited about. The possibilities are vast, and hopefully this tool will become a great asset allowing us to create better finished products.
Adriano Nunes is the associate creative director of Cognito sitting in our London office