Video created using OpenAI's SORA model/Official website, OpenAI
A week ago and without prior preparation, the OpenAI company released a website demonstrating the capabilities of its text-to-video model, Sora.
The world, and especially the world of marketing, will not see the same after this.
It feels almost trite to talk about yet another AI "innovation," but we need to talk about Sora because, judging by the demos, it looks like it's planning to do to video what the company has already done to text.
The level of video AI until the launch of SORA was reasonable at best.
There are some well-known companies in the market such as Kapwing and Canva, which made it possible to produce short videos from a prompt, and then join several such sections together, but the whole process was long and cumbersome and it made it difficult for users to achieve good results.
Now, Sora devours the cards and brings a new quality never before seen in these realms.
The new model allows you to create videos of up to a minute, combine two existing videos, animate images and more, all through text instructions.
The company notes on its website that Sora "can create realistic and imaginative scenes from text instructions."
How realistic?
Judge for yourself.
The power of SORA
On the SORA demo site we found a short video of a cat pestering a woman trying to sleep.
Apart from the woman's movement which looks a little strange, there is nothing revealing here.
The cat acts like a cat, and the woman gets angry.
It is a total of 8 seconds of video, but such an innocent video shows the real capabilities of the model compared to other models.
There are no explosions or spaceships here, there is a legitimate situation that could come up on Reelz or Tiktok, but it looks really, really real.
More on the dangers of this later.
One of the demonstration videos of SORA./OpenAI model capabilities
The marketing potential of Sora
Imagine a personalized video ad appearing in your social media feed.
The video starts with a picture of you, taken in a familiar environment, such as a favorite coffee shop or a park where you usually run.
After that, the video presents a new product of a well-known company, while personalizing the product and its benefits to your needs and habits.
For example, if you are a running enthusiast, the video will focus on the water resistance capabilities of the new running shoes;
If you are a coffee lover, the video will show the rich taste of the new coffee beans;
Do you like gaming?
The video will show you in front of a new TV with all the features that gamers love and so on.
A travel company can use Sora to create a video that presents a personalized itinerary for a customer, taking into account their interests, budget and time available.
All this information already exists in the meta.
Connecting the capabilities of SORA to an information base like Meta's and you have a machine that can produce everything and almost completely eliminates the need for a production team - designers, content writers, video editors, directors, photographers, producers;
Even the campaign manager has less work because they don't have to think about creative every time.
Personalized videos can lead to a better customer experience and an easier path to conversion from there.
They are more relevant and will interest the potential customer personally, because he is in the center, and not some presenter or network star.
With great power comes great pike
As with any technological breakthrough, we must talk about the dangers.
Used maliciously, Sora can be used to create completely realistic videos of events that never happened, leading to the spread of misinformation and fake news.
For example, a political entity can use Sora to present an imaginary event in which a political opponent is seen where he should not be, or says things that harm his good name, using a deep fake that already works well today.
Remember the disturbing video of Joe Biden fondling the wife of the new Secretary of Defense?
So with Sora he can do a lot more than that.
In addition, the new technology can cause a step up in the field of Ransomware - ransom scams.
There is a well-known scam in the US, where a parent is called from their child's phone number, informing them that they have been kidnapped and asking for a ransom. The "kidnappers" create a sense of pressure and panic with the help of falsifying the children's voices, so that the parents will transfer the money as soon as possible in exchange for their child who has never been kidnapped. We've already seen what can be done with the little information we have, like pictures, Sora takes the possibilities to other and far more dangerous levels.
Where is the truth?
The more AI videos fill social networks, the more reality will blur and we won't know at all what is real and what isn't.
Although there is a policy of flagging AI content in some networks, and Meta is working on an automatic flagging tool, but in the interim situation we are in, people may lose confidence in their ability to distinguish between fake and reality.
It already happens with pictures.
A video, which can tell a whole story, is much more dangerous in this respect.
Of course, OpenAI states on the website that they will place the appropriate "fences" for the strong model and build it in a responsible way that will not allow hostile parties to abuse it, but when you have a tool that provides such good results and so quickly (apparently, the model is not out yet), there will be It is very difficult to put the genie back in the bottle.
It can be hoped that the model will help more positive things like marketing, compared to dark things.
Summary
The new model, Sora, has tremendous potential to advance the field of marketing and the entire world, but it is important to use this technology in a responsible and ethical manner.
Before it is released to the general public, awareness of the dangers involved in using Sora should be increased, and steps should be taken to prevent the spread of misinformation and fake news.
And for those who run to try to download the model, wait a second, it is currently only a purpose display and the model is available for internal use to a number of artists, designers and filmmakers just to get feedback.
Until it is released, I hope that OpenAI will take a few more months to sit on the model and make it safe for everyone, without thinking about the bottom line, certainly when we are approaching an election period in the US.
Or Kurtz is a reporter and content strategist and the owner of Light-Web.
More on the same topic:
AI
artificial intelligence
video
Videos