Can you imagine typing just one sentence and, within seconds, an expressive minute-long video appears in front of you with its expressive characters and its multi-level scenes which never appeared at all? This is no more a future imaginative fantasy; it is here and now. OpenAI release of Sora in early 2025, followed by the Stability AI release of Stable Video later that year, summarised by some as marking the Onset of a wave of excitement some felt and fear others did about what generative video would bring to journalism, politics, and culture. There has never been a more blurred distinction between genuine and fake media and these lines will be broadened even further with the next election cycle, wait until there.
The State of Text-to-Video: From Labs to Your Screen
Twelve months ago, text-to-video models were largely research demos languishing in whitepaper and behind-the-thrill beta invites. Today, they are fast finding their way to mainstream platforms. Sora is able to convert plain text to video clips of standard quality (resembling the mid-budget stock footage), one minute in length. Gen-3 alpha has found its place in the creative agencies as the Runway is said to have a higher temporal consistency and more natural motion compared to all previous generations according to the verge. The 2025 Creative Trends Report created by Adobe showed that more than 70 per cent of agencies are currently trying AI video tools; be it in using them to create storyboard or in an entire marketing campaign.
The amount of investment being made highlights just how serious this change is becoming. According to a recent report published by Bloomberg Intelligence, not only venture capital but also acquisitions by Big Tech players should push the generative video market to more than 10 billion between 2022 and 2027. It is a gold rush- and, as always, a wild one.
Education, Advertising, and Synthetic Storytelling
In education, text-to-video holds enormous promise. Consider the scenario where medical students observe automatically produced surgical simulations, or the simulation of on-demand creation of dynamic role-playing videos in learning languages. In previous periods this year, Coursera experimented with generative training modules through cooperation with ElevenLabs and Runway. Students gave feedback that they were 30 percent more engaged than with recorded lectures normally.
Advertising is another frontier. A few months ago, I advised a middle-sized clothing store that has started to use automated advertisements based on Artificial Intelligence and targeting local populations. In less than two hours they generated 50 video variations- something that would have been unimaginable without generative technology. The cost savings in itself can transform the nature of small brands fighting against the household names.
Meanwhile, independent film makers are exploring the limits of fiction films. In 2025, the Sundance short film Mirage of Memory was composed entirely using Runway Gen-3 and ElevenLabs voice synthesis and executed with such levels of emotionalism that it became a sensation. Although it was all fake, viewers shed tears at theater. This incident remaines with me. We are taking an age where memories made by replicating the work of memories can be as emotionally moving as the things that actually happened to us.
Deepfakes, Consent, and the Crisis of Trust
Naturally, all this has the flip side of the darkness. In March, a False, hyper-realistic deepfake of President Macron making an announcement about an economic policy that never happened went viral in Telegram, earning more than 2 million views before AFP Fact Check debunked it. The video was very convincing that even respected reporters were afraid to declare it a hoax at once.
Such dangers do not belong to the political sphere only. Consent is becoming a legal quagmire. To whom does an AI-created actor belong? What occurs should the face or voice of an individual that is scrapped of the publicly available social media be utilized in a training dataset without express licensure? The draft of the AI Act by the EU, as found by Politico, has provisions of watermarking and revealing synthetic media but it is still an immense task to prosecute.
Regulators are not the only people struggling with these questions. As a consultant, I have received calls of brands in a panic about reputational damages through AI impersonators. Trust which is in precarious supply is the last casualty.
“The New Frontier Is Philosophical, Not Just Technical”
Recently, an AI ethics researcher at the MIT Media lab, Dr. Lena Kwan, told Wired:
Text-to-video not only makes our reality questionable, but it also sells it as a commodity. The question is can we afford to live in a world of synthetic experiences?
I do not disagree with that at all. It was strangely nostalgic watching as Sora made a video of my childhood neighborhood, which it recreated out of training data, and a few keywords. And unsettling. This was the memory which I never had, but somehow a part of me still wanted to believe.
All of us will have to reconsider what authenticity is all about, thanks to this technology. What happens when the next generation of folks are not concerned about whether or not something occurred so long as it is emotionally real?
Conclusion: The Blurred Horizon
Generative video is the most amazing generator I have ever encountered before. And it is the most dangerous one as well. On the one hand, it is supposedly going to democratize film-making, training and marketing, on the other, it is going to amount to the loss of integrity of all that we see and hear.
We are on the edge of a precipice on which the concept of evidence may dissolve. I cannot be more coarse: we need to build fantastic norms of transparency and cultural practices today or we will soon exist in a pretty world were the saying, seeing is believing, becomes little more than an embarrassing anachronism.
Well, next time you turn on a video on the Internet, you can ask yourself the question, do you have any idea where this came up? As you know, perhaps you will not in a brief while.