OpenAI Shuts Down Sora: The Truth About 'AI Filmmaking'
When a company closes a tool that promised to revolutionize cinema, what does it really mean for filmmaking?
With a press release, OpenAI announced the closure of Sora, the generative artificial intelligence tool that allowed creating video clips from text and promised to “revolutionize film production.” At the same time, Disney cancelled a one-billion-dollar investment specifically on Sora.
Sure, there are still VEO3, Runway ML, Seedance, or LTX Studio on the market, but this could be good news for those who believe the human touch in storytelling is still fundamental.
SometAImes it can be useful in Filmmaking
I recognize that tools like Stable Diffusion, Z-Image, or Midjourney can be useful in the early phases of creating a moodboard, but personally I prefer to privilege platforms like Frameset or Shotdeck in the pre-production phase. Text-to-image tools are the last resort since I prefer to prioritize authenticity above all else. And in any case, it’s better to work with a concept artist than with Draw Things or the defunct Invoke AI.
It’s a different story for AI-powered tools like Magic Mask in DaVinci Resolve and Magnetic Mask in Final Cut, or transcription functions that help me do faster what I was already doing myself.
Audience Perception of ‘AI Filmmaking’
Some time ago, a Baringa survey caught my attention in which it emerged that 52% of respondents would prefer to watch a film, perhaps imperfect, but created by humans rather than one technically flawless but created by AI. And the percentage rises to 57% among Gen Z (born between 1998 and 2008), who theoretically should be the preferred target of big tech investing in AI.
Moreover, almost all respondents agreed that AI content must be explicitly labeled, and this, in an era of deepfakes and reality alteration, is a significant data point.
If the public prefers human imperfection, why does the industry keep pushing AI? Perhaps because the economic discourse is more complex than it seems.
“If you don’t use Generative AI, you’ll be left behind.” They said.
I never bought this concept. For me, it’s exactly the opposite: delegating everything to a dataset in a sense ‘atrophies creativity’ and becomes an obstacle to improving as creatives.
Furthermore, the excitement that comes from creating is missing: at one point I played around with Stable Diffusion locally (I prefer to avoid cloud-based tools as much as possible) and often got good results, but I wasn’t creating anything and I knew that was nothing more than a ‘puzzle’ made from other artists’ work who might not have authorized the use of their material to train various AI models. Sure, nice to look at but without real meaning and going out and shoot it’s way better.
Image Rights and Ethical Dilemma in Filmmaking
Recently I read that Val Kilmer will ‘act’ in a film for which he had been hired but to which he couldn’t participate due to his passing. His image and voice will be reconstructed by AI and his heirs will receive compensation for the ‘performance’.
This creates a precedent that could change the perception of AI use in audiovisual media: actors will ask for higher and higher compensation to grant their image rights, and this could make AI use less profitable than one might think.
Image rights protection obviously also applies to ordinary citizens: a normal person who recognizes themselves in a commercial or fiction product created by AI can rightfully claim image rights, so AI filmmaking could become too expensive to be sustainable, besides generating a series of ethical questions. So, hire real actors and actresses.
This creates a precedent: if today we pay Val Kilmer’s heirs, what happens tomorrow when an ordinary citizen recognizes themselves in an AI commercial? Image rights could become a prohibitive cost for the entire sector.
AI’s Environmental Impact: Is It Really Worth It?
It’s known that data centers managing the clouds on which various AI models run produce emissions in the air, require a lot of storage (in fact SSD prices are rising) and need water for cooling processes necessary to prevent machines from overheating.
In this sense, news that a data center built in Uruguay limited drinking water supplies for residents in surrounding areas made headlines.
Is it worth it? Is this what we should ‘embrace’? Can innovation be defined as such when it benefits the community, otherwise it’s just technological progress that advantages a narrow elite. Is the priority really to produce images and videos with AI instead of using it to help research for curing autoimmune diseases, find innovative solutions to preserve the environment, or address the housing emergency? I don’t think so.
What to Really “Embrace”.
There’s a difference between creating and generating. Between telling and producing.
The closure of Sora made me think about all the hours spent shooting, editing, making mistakes and correcting. About those stories that maybe no one will ever see, except me. And about those that instead, for some reason, touch someone.
Preserving the genuineness of storytelling means choosing to be present. To go out with the camera. To sit in front of Final Cut Pro or DaVinci Resolve and see your story take shape, little by little.
This is my approach. If you want to see it in action, my portfolio is here: https://www.michelangelotorres.net.