The recent release of Sora, OpenAI’s text-to-video software, promises to boldly redefine the way we approach filmmaking. With this tool, you can essentially “write a video” from simple text instructions, bringing us closer to the dream of creating footage as effortlessly as typing up a document. Although it’s still early days, the potential is clear: for now, it’s best suited to short artistic pieces, commercials, and experimental content, but it may well serve as a stepping stone for more ambitious productions down the line.

Having had the chance to experiment with Sora as an alpha artist, I can say it’s among the most advanced text-to-video technologies I’ve encountered. That doesn’t mean it automatically produces masterpieces. Like all generative AI, it’s easy to end up with mediocre results. Achieving something truly memorable requires creative skill, technical know-how, and a willingness to embrace a certain amount of trial and error.

A differenza delle altre Ai generative, Sora richiede risorse significative in termini di costi, essendo disponibile con l’abbonamento a ChatGPT pro di 200$ al mese, ed è dunque uno strumento alla portata di pochi. Il target di riferimento sembrerebbe composto per lo più da professionisti o aziende. Purtroppo un po’ tutti i software text-to-video sono costosi per via degli ingenti consumi (un piano analogo della concorrente Runaway costa 100$ al mese, ma senza ChatGPT) e questi costi sono comunque in linea con i software per professionisti disponibili sul mercato (la suite grafica Adobe ad esempio costa attorno ai 70$ al mese). Va inoltre riconosciuto che nonostante il prezzo elevato Sora rende più accessibile la produzione audiovisiva, perché i costi di produzione video sono in genere molto, molto più alti.

Unlike other generative AI tools, Sora demands significant resources and comes with a hefty price tag—$200 per month as part of a ChatGPT Pro subscription—placing it beyond the reach of most users. Its target users appear to be primarily professionals and businesses. Unfortunately, steep costs are a common theme in the text-to-video market, largely due to substantial computational demands. A similar plan from rival Runway, for example, runs about $100 per month, though it doesn’t include ChatGPT. These expenses are nonetheless in line with other professional-grade software tools (Adobe’s creative suite, for instance, hovers around $70 per month). It’s also worth noting that, despite the steep subscription fee, Sora still lowers the barrier to creating audiovisual content. After all, traditional video production costs are typically far, far higher.
 

Another key factor is the introduction of what some are calling “guardrails”—built-in limits designed to prevent misuse or ethically questionable applications. For instance, you can’t push the envelope with violence, nudity, celebrities, copyrighted material, or other sensitive content. Some filters are so strict that they end up blocking perfectly harmless images or prompts, even by their own conservative standards.

In my view, these guardrails are less about genuinely protecting people and more about companies shielding themselves from legal trouble and avoiding additional media panic—a recurring pattern every time a new technology emerges. As Amy Orben has described, media panic follows a “Sisyphean cycle”: when confronted with a new technology, society experiences a surge of fear and anxiety, only to realize later that these worries were exaggerated or even unfounded. We’ve seen it with the introduction of the novel—accused in the 19th century of undermining public morals—as well as with the telephone, radio, television, rock music, and video games. Despite history showing these concerns are often overblown, the belief that “this time is different” somehow always resurfaces.

In Sora’s case, these guardrails largely stem from fears surrounding deepfakes—fabricated images or videos considered threats to truth, democracy, and young people’s mental health. Yet, as sociologist Kirsten Drotner notes, tech panics often aren’t truly about protecting vulnerable groups. Instead, they frequently serve as a way for older generations to safeguard their power amid shifts brought about by younger ones. The anxiety over deepfakes also relies on the outdated notion that media and images directly mirror reality. As I’ve mentioned before, public opinion is shaped far more by narrative framing and source credibility than by the authenticity of visuals, as evidenced by recent U.S. elections. Sora’s guardrails reflect these anxieties and end up restricting creative expression more than preventing genuine abuses. What we need isn’t ever-tighter (and often easily circumvented) controls, but rather a technological culture that fosters informed and responsible use. Our motto should be: teach rather than forbid.

Another controversial aspect of technologies like Sora is their environmental impact and their effects on the job market. It’s well known that training and running generative AI models consume significant amounts of energy. However, if we compare this ecological footprint to that of traditional productions—physical sets, crew travel, props, and extensive 3D rendering—the balance might be considerably more favorable. Of course, widespread use of Sora isn’t sustainable yet, but for more targeted projects, this technology could actually help mitigate some of the audiovisual industry’s environmental impact.

On the employment front, the arrival of text-to-video technologies like Sora is poised to fundamentally reshape the audiovisual industry. Some professionals—particularly those involved in large-scale productions—may see demand for their skills dwindle as Sora optimizes costs and timelines. At the same time, though, these tools open up opportunities for smaller creators and independent studios, enabling them to produce sophisticated visual content that would have been unattainable with limited resources. Accurately predicting the precise impact on jobs is challenging, just as it was when desktop publishing software was introduced. Many feared that traditional roles would vanish entirely, but reality proved more nuanced: while some old roles faded away, new ones emerged.

The real risk lies in allowing economic considerations alone to guide the adoption of these technologies. Reducing Sora and similar tools to mere cost-cutting measures would be shortsighted. The true potential of AI lies in its ability to broaden creative horizons, offering fresh languages of expression and new possibilities. Yet life under capitalism often means innovation is subordinated to profit, and the gains of a few can overshadow those of the many. Once again, the problem isn’t the technology itself, but the social context that shapes—and is shaped by—it.

The recent Sora leak, orchestrated by some artists involved in the testing program, released confidential materials before the official launch and was accompanied by a manifesto critical of OpenAI. The authors of the leak protested the lack of compensation for their work and accused the company of imposing undue creative constraints through its guardrails and content controls. They ended with a call to use open-source software.

While I also believe that all generative AI technologies should be open source—freely available and modifiable without restrictions—and I share many points of the manifesto, something rings hollow. First and foremost, the manifesto paints us alpha artists as naïve pawns of a multinational corporation, yet in my view, the conditions offered by OpenAI for testing Sora were more than fair. We had free access to the software, no obligation to provide bug reports or promote the company (otherwise, I wouldn’t have accepted), and the freedom to publish what we created as long as we respected intellectual property rights. It was a good deal—an opportunity to work ahead of the curve with a remarkable tool, gaining a clear edge over other professionals.

The “rebel” artists also enjoyed these benefits for months, only pulling out a few weeks before the official release. Sure, the prizes (around $1,500) for the internal contest were low, but participation was voluntary, and I personally took part. For me, it’s not just about the money. As someone who isn’t that famous, exhibiting my work alongside other talented artists, traveling the world at OpenAI’s expense, and being associated with a leading AI company is excellent publicity and exposure.

In short, the reason why I—and many others—can’t endorse the call to action is simple: I can’t afford it. Even the letter’s signatories acknowledge that open-source software requires substantial hardware and skills that not every artist has or wants to acquire. Working with open-source solutions offers many advantages—such as privacy and freedom from censorship—but taking it to a professional level remains out of reach for most. These tools aren’t “gatekeeping-free” at all. For me and many others, following the manifesto’s advice would mean simply giving up on AI, which contradicts the very openness and inclusivity the advocates claim to champion. Some artists just don’t have the means. Given the cost of producing video content, if you lack technical expertise or powerful hardware, subscribing to Sora for a month or two ($200–$400) to fund a video is currently the most practical solution.

I believe we should focus the debate on making these technologies as accessible and transparent as possible, without creating new technical or economic elites. We need to push companies for more freedom and transparency and, realistically, lower prices or provide scholarships and discounts for students and less affluent professionals. And if we really want to look beyond our own backyard, we should ask OpenAi not to make its technologies available to the military.
We should also urge legislators to strongly support the development of public, open AI. The European AI Act was a missed opportunity: if we had freed open-source projects from copyright obligations, we could have turned Europe into the world’s foremost open-source hub.

Generative AIs like Sora are at a stage reminiscent of personal computers or other technologies: amazing tools, expensive to develop and produce, and somewhat accessible to the public—but still behind a financial gate. Our challenge is to tear down—or at least lower—that barrier.