24th October 2025 | The Friday Filmmaking Blog

It’s an AI triple bill this Friday. We’ve got CineD’s sharp take on Sora 2 and why Hollywood’s gone to war over copyright and consent; The Guardian’s on-set investigation into body scanning and “AI doubles”; and No Film School’s coverage of Kevin O’Leary’s claim that A24’s Marty Supreme should’ve swapped human background artists for AI extras. Together, they sketch a clear fault line for filmmaking in 2025: dazzling capability on one side, and a scramble for legal, ethical and creative guardrails on the other. If you care about where filmmaking goes next, this week’s reads are essential.
Sora 2 vs Hollywood – The Copyright Reckoning of Generative Video
CineD’s deep dive into OpenAI’s Sora 2 captures both the wow-factor and the whiplash. On the one hand, Sora 2 sets a new bar for consumer-accessible text-to-video: more realistic physics, better continuity, and integrated audio, all packaged in a free iOS app that briefly rocketed up the App Store. On the other hand, its launch triggered the fiercest copyright backlash the industry has seen since early YouTube. CineD lays out why: the app initially allowed copyrighted characters and likenesses to appear in outputs unless rightsholders opted out—effectively reversing the default of consent that underpins modern IP law. That design choice supercharged virality (SpongeBob flipping burgers; Mario in a starfighter) and supercharged Hollywood’s response.
Trade bodies and talent companies moved fast. The Motion Picture Association publicly demanded action; top agencies reportedly told clients to opt out; SAG-AFTRA warned the model “threatens the economic foundation” of the industry. For studios, CineD notes, the core dispute isn’t just training data, it’s finished outputs that reproduce distinctive characters and likenesses—a realm where infringement risk is clearer and E&O insurers are already skittish. In short: you can’t build a reliable production pipeline on a tool if you can’t guarantee clean chain-of-title.
Within days, OpenAI signalled a pivot. CineD cites CEO Sam Altman’s “Sora Update #1,” promising more granular rightsholder control, a tilt toward opt-in, and even exploring revenue-share models that frame Sora as a licensed “interactive fan fiction” platform. PR framing aside, the article argues that the damage was done—copyrighted content had already been generated and widely shared under the app’s initial rules, and outrage over “AI slop” clips only intensified the mood.
CineD also zeroes in on performer rights. Sora 2’s “cameo”-like ability to insert real people into generated footage edges into likeness rights—terrain that unions fought hard to define in recent contracts. It’s a reminder that AI in filmmaking isn’t just about shots and shaders; it’s about consent, authorship, and who gets paid when a face (or a character) appears on screen.
For working filmmakers, the nuance here matters. The tech is undeniably useful for previs, pitch films and indie-scale ideation. But until licensing and liability are nailed down, deploying Sora 2 in commercial delivery is gambling with legal exposure. CineD’s takeaway is practical: experiment, learn, and keep your ethical house in order—transparency, consent, and authorship are now core craft skills alongside lighting and lenses. And if you want a cultural pulse-check, the piece points to a timely commentary video that captures how quickly this moved from “fun toy” to industry flashpoint.
Have we done ourselves out of a job?’: concerns in film and TV industry over on-set body scanning
If CineD maps the platform fight, The Guardian maps what’s happening on set. Michael Savage reports first-person accounts from actors, stunt performers and dancers who are being ushered—sometimes at the end of long days—into full-body scanning rigs with opaque explanations of how their biometric data will be used. The anxiety isn’t theoretical: those scans can seed “AI doubles” that live beyond a single production. Supporting artists (SAs) describe hearing crew openly speculate that crowds can now be conjured from stored data, potentially shrinking headcount across ADs, hair and make-up, costume, catering and drivers. That ripple effect is the real bombshell: the risk doesn’t stop at performers.
Actor Olivia Williams offers one of the piece’s most striking testimonies. She describes the social and professional pressure: you’re mid-scene, an AD you trust asks you to “pop over” to the VFX bus, and the unspoken message is clear—refuse and risk future work. Williams argues that performers need scan controls with the teeth of a “nudity rider,” i.e., explicit opt-ins for context, scope and duration. The article also notes how quickly the Tilly Norwood saga—the widely publicised “AI actress”—has become a lightning rod, focusing wider unease about identity and substitution.
Legally, The Guardian spotlights a vacuum. AWO lawyer Alex Lawrence-Archer describes overlapping, outdated contract language that was never designed for this technology. In that grey zone, studios and AI vendors push as far as they can; performers who try to claw back control discover the expense and ambiguity of asserting data rights after the fact. The recommendation is pragmatic: nail the terms up front. Some casts are now forcing addenda that lock scans to a single show and forbid reuse without fresh consent—a tiny but important wedge of control.
For filmmakers, this piece is a wake-up call on process design. Scans aren’t just a VFX utility; they’re a consent workflow. Build time, documentation and clear language into your call sheets. Define retention, reuse, and destruction policies. Budget for human crowds where authenticity matters, and be honest when synthetic augmentation is being considered. The cultural optics are volatile: in a year of strikes and AI anxieties, a casual, after-hours scan can read as exploitative—especially when it’s captured under fatigue. The upshot for filmmaking is simple: the more transparent and specific you are about scanning, the more trust you retain with cast and crew, and the less legal risk you carry later.
‘Just Put 100 Tilly Norwoods In There’: Kevin O’Leary Says Movie ‘Wasted Millions’ on Human Extras
No Film School covers a headline-grabbing claim from investor-turned-first-time-actor Kevin O’Leary: that Josh Safdie’s A24 drama Marty Supreme “wasted millions” hiring background artists and could have saved a fortune by using AI extras—“just put a hundred Tilly Norwoods in there.” The comment came during a podcast appearance while O’Leary was promoting the film, and it leans into a certain Silicon Valley logic: background work is “only visual,” therefore ripe for synthetic substitution. NFS quotes him suggesting the director could have made two films for the cost of one if AI crowds were allowed.
The article pushes back from a filmmaking standpoint: period authenticity with 100–150 bodies in frame isn’t just pixels, it’s texture—wardrobe interaction, eyelines, ambient behaviour and the ineffable “lived-in” quality that extras deliver. The piece also points out that O’Leary repeatedly garbles the name of Tilly Norwood, the controversial AI “actress,” undercutting the authority of his prescription even as it illustrates a broader shift: the AI debate is now happening in budgets and press tours, not just labs and forums. For producers and UPMs, that means you’ll hear more of this argument in notes meetings and boardrooms. Be ready with craft-led counters, not just ideology.
Context matters. Tilly Norwood’s rise crystallised concerns about AI characters trained on human data and packaged for commercial use; unions and talent agencies have been vocal about the consent and compensation gaps. Swapping real SAs for synthetic avatars doesn’t just change a line item—it redistributes income in an already precarious ecosystem, echoing the Guardian piece’s warning about knock-on job losses across departments. Practically, there are also legal and PR risks: likeness scraping, undisclosed compositing, and the optics of replacing workers mid-contract. In a brand-driven indie like an A24 film, that blowback can be costlier than the background day rate you “saved.”
Still, NFS acknowledges the market reality: pressure to rationalise costs won’t vanish. The smart path is situational use of AI—crowd tiling for far-background set extensions, previs to plan blocking, or synthetic augmentation where faces are genuinely indistinct—paired with explicit consent frameworks and honest crediting. That approach keeps the creative integrity of filmmaking front and centre while insulating productions from the reputational damage that comes with wholesale replacement rhetoric. Close the read by watching the official teaser—and imagine how quickly the period spell would break if the background crowd behaved like a row of cloned NPCs.
Latest News
Thinking about using an award winning production company for your next project?
Want to chat through a video or animation idea you may have?
Planning on using video or animation in your next campaign?
We’d love to hear from you. Get in touch today!


