CONVERGER’S AI in Hollywood News Roundup
All the news that's fit to stack about AI's collisions with the film business in July and August
Welcome to CONVERGER, a biweekly(ish) newsletter mapping the content singularity where AI and the internet collapse all media into one—a connective node where emerging technology, policy, culture, futures thinking and storytelling intersect.
I’m Kevin Bankston, your host. You can watch me develop newsletter content in real-time on LinkedIn and the social network formerly known as Twitter, and less often on Bluesky and Instagram.
In today’s issue, we’re mapping the signals of AI’s accelerating convergence with film and TV. (Some of you may have seen an earlier, less comprehensive version of this roundup posted earlier as a note; this one is much bigger and better!)
TABLE OF CONTENTS
Studio AI Adoption: Signals and Scrutiny (~1100 words, ~5 min read)
Filmmaker Dialogue and Developments on AI (~650 words, ~3 min read)
And in Other News: Tools, Trends, Tidbits (~1150 words, ~5 min read)
Studio AI Adoption: Signals and Scrutiny
Since our last major round-up of news in early July, there have been an enormous number of developments at the intersection of AI and Hollywood. That includes some big stories about how Hollywood studios are quickly adopting generative AI into their production workflows, with some being forthright about it and others hiding it (and then getting found out).
Netflix in its quarterly earnings call revealed that AI-generated imagery has been used in over 300 programs on the platform in 2026 alone, highlighting it as a cost-saver. The fact that no one really knows which shows have used the technology, other than the few like The Eternaut that Netflix has previously publicized, suggest relatively subtle uses in special effects, animation, and backgrounds. But I think this is basically the starting gun for broader use across Hollywood.
Another production company that isn’t shy about using generative AI is the Wonder Project, which produces TV series House of David and The Old Stories: Moses for Amazon using a hybrid production model of human actors in front of generative-AI backgrounds on an LED volume screen. The director of those faith-based shows, Wonder founder Jon Erwin, also directed the new feature film Young Washington just released in theaters by Christian film distributor Angel Studios, and like Erwin’s TV shows it makes extensive use of generative AI which he was happy to detail in an interview with Variety.
Meanwhile, at the major studios: MIT’s Sloan Management Review published a case study about gen AI’s integration across Warner Bros. Discovery, both in business and production processes. It noted a strategic difference from Netflix: WB leadership aligned on focusing on using AI for growth rather than efficiency, including avoiding using the technology to cut jobs. Key uses in production focused on metadata generation, marketing content generation, localization initiatives, and animation applications, including storyboarding, ideation, and establishing backgrounds and shots that would have been prohibitively expensive otherwise, while staying away from using it for primary character animation due to consistency challenges.
Two excellent LA Times articles from crack AI-in-Hollywood reporter Nilesh Christopher (Nilesh — Out of Distribution) further highlighted AI’s spread. The first focused on how the video model Seedance from TikTok parent Bytedance is increasingly being used in productions, even if sometimes with a “don’t ask, don’t tell” wink and nod, noting that Bytedance has been offering studios special unlimited access for $2 million. (As part of its warming relationship with Hollywood, Seedance 2.5 has also stepped up the aggressiveness of its copyright and likeness filters for its consumer product, apparently tied to a new agreed-upon framework with the MPAA, although whether those filters would apply to the studios’ all-access deals is unclear). Second was an examination of the major studios job listings in late June, highlighting how more than 1 in 10 was related to AI, with those jobs clustering at the studios that are natively tech companies: Amazon MGM, Apple, and Netflix.
Echoing the controversy last month when the director and distributor of Stop! That! Train! admitted, denied, and then admitted again to having used gen AI in that movie, horror director Eli Roth is now saying he misspoke when he suggested that AI-generated animated sequences in his new pic Ice Cream Man were fully human-authored. Then that AI controversy was swiftly followed by another: a behind-the-scenes video for Supergirl appeared to show the director Craig Gillespie and actor Jason Momoa looking at a selection of concept art of Momoa’s character Lobo on a laptop that some folks online immediately concluded was AI-generated, with consequent freak out. To be clear, it’s not obvious the images were AI-generated or the application they were looking at was an AI generator. Even if it was AI, it’s not clear they didn’t use an “ethically-sourced” fully-licensed model like Adobe’s Firefly or Moonvalley’s Marey, and perhaps more the point, they were generating images of their character and their actor’s likeness, so cries of “IP theft” or “plagiarism” seem pretty unsupported. Which leaves at most the objection that this might possibly put some concept artist out of work, though we know from the Warner Bros. Discovery case study (Supergirl is a Warner’s property) that’s not their strategy. So, seems like a tempest in a teapot to me—but Warner pretty much invited more criticism when it took the video down instead of clarifying whether or how they’d used AI. They perhaps could use a little more of Netflix’s spine in saying what kind of AI future they want to build or not, especially since—as Puck’s Matthew Belloni said in a recent interview with Wired about AI use in Hollywood—”Everyone’s doing it.”
Then Warner Bros. needlessly stepped in it again with another AI controversy when promoting dinosaurs-invade-the-suburbs pic The End of Oak Street, having the family dog from that movie do a podcast interview segment with the AI-generated talking dogs from “The Dogbone Podcast.” I know they say that no publicity is bad publicity, but I’m shocked that anyone on the marketing team thought this was a good idea considering the current environment around AI in Hollywood. (Side note: the movie is great! Go see it so the studios will make more like it!)
FWIW, all of these developments are making me rethink my position on AI labeling. Speaking generally, my take is Duke Ellington’s, who would say when responding to people who complained that jazz wasn’t “real” music because it didn’t follow traditional musical rules: “if it sounds good, it is good.” If a work speaks to me, I don’t tend to care how it was made. But clearly a lot of people—or, at least a very vocal minority on Twitter—do care.
I recently saw a metaphor articulated in a Substack note that resonated with me: food labeling. Sourcing of your food may matter to you regardless of if the taste is the same, and I think consumers who care should have a choice to support one type of source over another.
Of course, the hard question is how to draw the right lines in terms of what and how to label (e.g., AI-generated vs. AI-assisted vs. AI-edited vs. AI-something-else) and exactly how detailed or prominent those disclosures ought to be, and I certainly don’t think the government could mandate labels without a serious First Amendment problem.
But I’m coming around to voluntary labeling frameworks as a worthwhile and desirable mitigation to help human creators and creations attract the audiences that want to support them, particularly in this early and chaotic phase of our adaptation to AI diffusion. At least, that’s what I’m thinking this week. I’ll be taking a deeper dive into the issue of labeling in a future issue, and reserve the right to change my mind!
Filmmaker Dialogue and Developments on AI
The OG tech-forward filmmaker, George Lucas, said ““Artificial intelligence means it’s much easier for us to make movies,” likening ignoring it to preferring the horse and buggy over the car. “That’s progress, it’s the future.”
Fellow tech-forward director Darren Aronofsky is seeking to raise $15 million in equity for his AI studio Primordial Soup, which is also hiring for a head of studio and multiple generative artists. The studio has a previously-established strategic partnership with Google Deepmind with which it produced the AI-generated historic shorts miniseries “On This Day… 1776” for TIME.
Another sci-fi director, Neill Blomkamp (District 9), posted a new AI-generated short film, zombie soldier action thriller Nightborne, from his new AI studio Barley. It was...not well-received online. On the bright side, it employed real concept artists and used fully licensed likenesses and voices of 32 performers.
In contrast to his elders, young Backrooms director Kane Parsons is going in the opposite direction: “If I could snap my fingers and make generative AI disappear forever, I probably would…. To me, generative AI feels less like innovation than a symptom of a broader cultural and economic rot.”
Parsons’ perspective is one proof point for The Odyssey director Christopher Nolan’s rather grand assertion that Gen Z is anti-AI: “I’ve never seen a more rapid wholesale dismissal of a supposedly foundational jump in technology in my lifetime…. So much energy has been expended on bringing in AI, but if you look at that generation’s reaction, they’re utterly rejecting it.”
Speaking of The Odyssey: Elon Musk is bloviating about making a more “historically accurate” (read: white) The Odyssey using Grok by the end of the year. Like most things he promises, it will likely never happen or take ten times as long as he predicted. Meanwhile, Ash Koosha, director of the first AI-generated film accepted into a major festival (Tribeca) and then roundly panned, has announced he is now directing his own Odyssey fan-fic, Odysseus: The Fall, using Kling AI. The Ankler puts this trend in context: expect to see a lot more low-budget AI-generated “mockbusters” following up and drafting off the marketing for bigger blockbusters, as well as more AI-generated slop adaptations of public domain works.
Hit AI creator Gossip Goblin aka YouTuber Zack London will be debuting Gods Don’t Give Gifts, a feature length anthology of his work, in theaters on October 30. Bit of an odd choice since anthologies are renowned for regularly failing at the box office, and Gossip Goblin’s sci-fi is already pretty weird and disturbing for a general audience. But this sounds more like a limited run targeted at his online fans rather than a broader audience play, so we’re still waiting for the first real broadly-targeted AI-generated feature film wide release.
AI studio Promise announced it was developing an AI-generated horror feature, Touch Grass, with Promise’s Dave Clark directing and an EP from Backrooms producing. Its premise involves a hidden world of bots underlying the internet, building on the “Dead Internet Theory” that suggests that most internet content and activity is now from bots rather than humans. Unlike Higgsfield’s features and like Jon Erwin’s productions, this one will be a hybrid production combining “live-action actor performances, practical effects, traditional visual effects, and generative AI.” Promise’s comms team has been working overtime since they’ve been everywherre: there was also Hollywood Reporter piece about Promise bringing in traditional TV directors and producers who’ve worked on shows like Doctor Who and Sharknado to collaborate with their generative artists, a Forbes feature on Dave Clark, then a Guardian spotlight on the studio tied to the Touch Grass announcement.
Finally, one more AI creator announced he was directing an upcoming feature: PJ Accetturo aka PJ Ace is partnering with Dreamina AI for his upcoming fantasy flick Nexus, which will be also be shot hybrid on a volume stage.
And in Other News: Tools, Trends, Tidbits
We finally have a hard number: Netflix revealed that Ben Affleck got $587 million from the streaming giant for his AI startup Interpositive; Affleck and Netflix were cagey about the exact amount when the deal was announced. As highlighted in the very first issue of CONVERGER, they were also less than forthcoming about whether and how Interpositive’s models relied on unlicensed training data, implying they did not. But its patents told a different story…
Speaking of unlicensed training data: video modelmaker Midjourney, defending itself in a lawsuit by the studios for training on their content, hopes the judge agrees that turnabout is fair play: they are asking for discovery into how the studios themselves are using AI internally, arguing that “[i]f Plaintiffs are doing the very thing they seek to punish, that evidence goes to the heart of Midjourney’s fair use and unclean hands defenses.”
The replacement of human actors in Chinese microseries continues apace, and audiences don’t seem to mind. A fascinating piece by Erik Barmack at The Ankler Team reveals that in January, 38 of the top 100 short-form shows in China starred AI actors, and that more than 95% of the approximately 128,000 new Chinese microdramas released in the first quarter of 2026 were AI generated. The US is still far behind the curve on microdramas compared to Asia, but there’s no reason to think this trend won’t eventually reach our shores, especially as major Chinese microdrama producer COL Group is expanding its global reach. I’ve always been skeptical that people would care to knowingly watch fake people in their stories, but that may have been wishful thinking…
In the world of Hollywood microseries, a concerning development: actors are seeing their filmed performances for the vertical dramas being overwritten by AI avatars.
Turkish TV series Castle Walls debuted on Amazon Prime last Wednesday, in what is being touted as the first fully AI-generated TV show to premiere on a major global streaming platform worldwide. Not much of a series—apparently just two 15-minute episodes—but you can watch it here.
Tech reporter Taylor Lorenz wrote a 2500-world deep dive into her experience with gen AI toolmaker Higgsfield, the self-proclaimed fastest growing AI startup in history, and it’s an unflattering portrait of shady upselling, predatory pricing, and disappointing video results.
More flatteringly, Higgsfield took the interesting step of open-sourcing all of the assets and prompts behind its 95-minute fully-generated feature film Hell Grind, which notoriously screened in Cannes but not at Cannes. It also released a handy-looking PDF guidebook to building an AI-generated feature film pipeline, tied to its recently-posted next AI-generated feature, action comedy caper Cully Hill Boys, which as reported by Variety notably features characters based on the licensed likenesses after the streamer N3on, UFC fighter Israel Adesanya, actor and UFC fighter Quinton “Rampage” Jackson, and basketball-focused influencer Matt Kiatipis.
More Higgsfield-flattering: legendary Pixar cofounder Ed Catmull joined the jury for Higgsfield’s upcoming AI film festival. More Higgsfield-unflattering: you have to be a Higgsfield subscriber and make your film on the Higgsfield platform to enter.
The Hell Grind assets release and Cully Hill Boys PDF aren’t the only interesting development in AI studios and toolmakers releasing educational materials. Also check out this new three-episode course from Magnific walking through the creation of a short animated film using Seedance 2.5.
More toolmaker news: Foundry is launching an enterprise version of its Griptape workflow for studios to run completely internally on their own infrastructure, to better secure their proprietary models and data. Also targeting studios, startup Copysight is pitching a system that can give AI outputs a copyright risk score.
Bytedance-owned editing toolmaker Capcut is inviting applications for its Capcut Creative Partner Program, “an invitation-based community for filmmakers, storytellers, AI creators, and creative professionals who are pushing the boundaries of content creation with CapCut.” Members get free credits, free Pro membership, early access to new models and tools, and other perks.
The 2026 cohort for the Disney Accelerator program, now in its 12th year, is exclusively AI companies, including gen AI toolmaker OpenArt and gen AI studio Promise (they are everywhere, apparently).
Speaking of proliferating AI studios, Forbes published an interesting roundup comparing and contrasting the IP strategies of the many different emerging AI production houses.
At the Metrograph theater in Manhattan, Google just showcased the films from the fourth cohort of directors to go through its “Flow Sessions” program, a six-week intensive that gives a group of aspiring filmmakers unlimited access to Google’s Flow short-form AI video generator paired with mentorship to help them use it. Variety headline editors breathlessly ran with a drug metaphor: “Inside Google’s Efforts to Get Creatives Hooked on AI.”
The new open source multimodal video generator MinMax H3 just dropped, big enough to have serious generative horse-power but also small enough to run locally on your own computer. AI Twitter is going crazy making videos locally with no IP guardrails and mashing up every conceivable IP, while the Hollywood press is going crazy about AI Twitter making videos locally with no IP guardrails and mashing up every conceivable IP.
Speaking of open source video models, Black Forest Lab’s much anticipated multimodal Flux 3 model finally dropped, with open weights to come soon.
Also on the open source tip, AI creator Sam Wasserman has been doing the whole AI community an enormous service by vibe-coding an entire toolkit of open source AI video production tools as an alternative to subscribing to companies like Higgsfield. You can support his effort here.
Finally, not AI-related, but Substacker Daniel Best tells a profoundly interesting tale of superhero movies, comics and copyright in his new article DC vs Wozniak: Who Owns The Copyrights To Unpublished Works?, about comics writer Christopher Wozniak who long ago pitched an idea to DC for a Batman story that they didn’t buy and then saw some of its broad strokes repeated in The Batman movie. The court never reached the question of whether the movie copied Woz’s story pitch because it concluded that the pitch itself was an infringing derivative work by virtue of using DC’s characters. This would be a surprise to most working freelance comic writers who regularly write pitches for Marvel and DC for those companies’ characters, and when rejected use the same plot idea in their later work using different characters (which under this court’s reasoning would be a further infringement). This routine state of affairs was described at length by noted Batman comic writer Chuck Dixon in one of the depositions in the case which was excerpted in an earlier article from Best about comics pitching, but based on this decision writers should probably not rely on that status quo anymore, and instead make sure their asses are covered in some sort of agreement before they pitch ideas using corporate characters.
And that’s all the AI in Hollywood news that’s fit to stack, for now! See you next time.