In May 2026, Belief Studio presented its first publicly showcased AI-augmented film work. The film Here’s To You, My Son speaks about healthy masculinity and did not originate from a brand brief. No client paid for it. The studio financed it itself.
In the same period, the opposite story was playing out in Hollywood. Lionsgate, which has a library of over 20,000 film and television titles, signed a deal with Runway in September 2024 for its own AI model. A year later, The Wrap reported that the partnership had not yet produced any publicly credited production built on Runway’s custom model for Lionsgate.
That is a useful contrast. A large IP library is not the same thing as a production workflow. And a small team with a clear creative brief can use AI to get out a film that would never have been made under a classic production budget.
That difference matters, because AI production is not one discipline. Several things run side by side: in-house custom models for studios, commercial generative tools for marketing assets, licensed models for film compliance, and AI-augmented production pipelines at smaller creative studios.
Belief Studio turns AI into a tool for a human brief
Belief Studio was founded in February 2026 within the structure of the London production company FORM. Its founder and creative partner, Elliott Starr, formerly of Impera, 20something, Fallon and Leo Burnett, launched the studio together with FORM’s managing partners Dave Kennedy and Jonathan Harris.
Kennedy described the model as a combination of “the scale of new technology with the artistic taste and craftsmanship of great human minds”. It operates inside FORM, but also takes its own commissions from agencies, creative companies and brands.
The studio’s first film, “Here’s To You, My Son”, came out in early May 2026. Parents and grandparents from Britain answer the question of what it means to be a good man. The testimonies are human. The visual execution and post-production composition rest on AI.
Starr explained the reason for the production choice in these words: “the short answer is budget, we didn't have the budget to shoot it, we'd need 27 actors in 13 locations across the country”. No client paid for the project, the studio financed it itself.
Paul Frampton-Calero, CEO of the consultancy Marketing Transformation, commented on his LinkedIn reshare of 18 May 2026 with the words “you were able to leverage AI prompting to create this message when no brand is brave enough to fund it themselves”.
Starr has summed up his method repeatedly on LinkedIn. About the film on masculinity, on 13 May 2026 he wrote that “hitting 'generate' is the least interesting part of the process. It's what precedes and follows that matters, the same as a camera. Oscar winners and bargain-bucket B-movies often use the same equipment.”
In a post from 20 May 2026 he drew an analogy with the camera. “Anyone can go online and pop a camera on Klarna, so why isn't everyone Erwin Olaf, or Lilly Wachowski?” Access to a tool and the result are not the same thing.
Lionsgate and Runway a year after the deal
Lionsgate and Runway announced the partnership on 18 September 2024. It was the first deal between a Hollywood studio and a generative AI company. Runway was to train a custom model on the Lionsgate library for previsualisation and visual effects.
Michael Burns, Vice Chair of Lionsgate, spoke of the possibility of saving “millions and millions of dollars” in pre- and post-production. Cristóbal Valenzuela, CEO of Runway, framed the deal in the words “the history of art is the history of technology”.
A year after the signing, The Wrap reported on 22 September 2025 that Lionsgate had not yet presented any publicly credited production on its own model. An insider quoted in the article described the limit in these words: “The Lionsgate library is too small to make a model out of. Disney’s library is too small as well.”
On top of that came questions about the rights of actors whose faces and likenesses are captured in films. A Lionsgate spokesperson said that the studio “continues to pursue AI initiatives according to plan” and that the Runway deal “is not exclusive”. In practice Lionsgate uses tools from several suppliers at once, and the custom model has so far not delivered a specific film that the studio publicly credits to Runway.
That is a useful benchmark. A studio with annual revenues in the billions of dollars and a library of 20,000 IP titles has not turned its AI model into a finished film in a year. A small AI-augmented studio inside FORM, without brand funding, released an emotional short film at the same time. What matters is what the team prepares before pressing the Generate button.
Klarna counted its savings, Coca-Cola took the criticism
Brands that deployed AI directly in advertising production have already supplied concrete data. In a press release of 28 May 2024, Klarna reported annual savings of around $10 million in marketing, of which $6 million came from image production and $4 million from external suppliers.
Spending on marketing agencies fell by 25%. The development cycle for image creative shrank from six weeks to seven days. David Sandström, Klarna’s CMO, described the deployment of Midjourney, DALL-E, Firefly, Topaz Gigapixel and Photoroom in these words: “AI helps us be leaner, faster and more attuned to what our customers care about”. Klarna applies these figures across all 45 of its global markets.
In November 2024, Coca-Cola went into an AI Christmas advert as a modernisation of its 1995 “Holidays Are Coming” campaign. The spot drew criticism from audiences and the creative community for giving an impression of a cheap shortcut, an uncanny visual and the replacement of human production work. Adweek documented the negative reaction across the creative industry at the time.
In 2025 the brand returned with another AI holiday advert. Pratik Thakar, global VP and head of generative AI at Coca-Cola, then argued that “last year people criticized the craftsmanship, but this year the craftsmanship is ten times better”. Jason Zada, founder of one of the supplier studios, Secret Level, framed part of the criticism as resistance from a creative industry that is afraid for its jobs.
For a brand of this size, an AI execution is easily read as a cheap shortcut, unless strong creative direction is visible behind it.
Asteria and Marey build on licences
Asteria, the AI animation studio of Bryn Mooser and Natasha Lyonne, launched the Marey model in partnership with Moonvalley on 8 July 2025. Moonvalley and Asteria build Marey on the claim that the model is trained exclusively on licensed data, with Moonvalley signing multi-year contracts with provider partners and producers who gave their consent.
Business Wire described the model as one of the largest video AI models built exclusively on clean data. The price for creators is $14.99 for 100 credits, $34.99 for 250 credits and $149.99 for 1,000 credits. Marey offers text-to-video, image- and video-to-video, fine-tuning on your own data, camera control and object motion guidance.
At the same time, Runway has moved its product from a single tool to a broader platform. Its current pricing has five tiers from Free to Enterprise, and the paid plans include Runway’s own video models as well as access to selected third-party video and image models, such as Seedance, Kling, FLUX and Seedream.
Marey and Runway diverge in the market on a key point. Marey builds its argument on licensed data, which Moonvalley and Asteria offer, according to TechCrunch, to studios and brands with strong compliance features. Runway builds its argument on an ecosystem of models from which creators choose. Belief Studio does not fit into this map, either as a model company or as a pure tool provider. It sells a specific craft, the concept and the ability to use AI in the production process.
What the IAB reports on US digital video
The market these tools and studios are aiming at is growing. In its report of 5 May 2026, the IAB estimates that US spending on digital video advertising will exceed $80 billion in 2026, growing 11% year on year, which is almost a fifth faster than the total advertising market.
Social video grows 13% year on year and for the first time overtakes CTV, which grows 11%. The IAB attributes this to AI-driven personalisation and higher spending on the creator economy. Geographically, the data covers the US only.
Klarna uses its AI budget for higher output frequency and lower unit cost. Coca-Cola used it in December 2024 for a big holiday campaign and, according to Adweek, faced feedback that Pratik Thakar and Jason Zada both played down in interviews.
Belief Studio uses it for a film that would otherwise not exist, because no one would have financed it as a classic production. Each case has a different line in the ledger and a different measurable result. The common denominator remains a lower cost of creating the visual output.
What the EU requires from August
Article 50 of the EU AI Act takes effect on 2 August 2026. Providers of AI systems that generate synthetic image, text or audio must, according to the European Commission’s guidance, ensure machine-readable labelling of content.
Deployers, that is, entities that deploy or publish AI content, must disclose that the material is artificially generated if the result appears to be purely human-made content. For generative AI systems placed on the market before 2 August 2026, a grandfathering rule applies, with a watermarking obligation by 2 December 2026.
For Belief Studio, an exception applies to “evidently artistic, creative, satirical or fictional works”. The disclosure must be made in a way that “does not interfere with the perception of the work”.
For Klarna ads that are meant to look like a real product photograph, the usual disclosure obligation applies. For a similar AI-generated advert distributed in the EU, a regulatory risk would arise from August 2026 if the brand did not clearly disclose to consumers that the content is AI-generated or AI-manipulated. For breaches of the transparency obligations under Article 50, the stated sanction framework is up to €15 million or 3% of global annual turnover, whichever is higher. A code of practice on transparency, which will clarify these rules, is expected in June 2026.
What determines AI production today
In AI production today, workflow, rights, creative direction and the ability to finish and release the result decide. The size of a player does not guarantee these things on its own.
The Lionsgate and Runway case shows that a large IP library alone does not produce a finished film. Klarna showed that AI can lower unit costs and speed up marketing production. Coca-Cola showed how quickly, for a major brand, an AI execution is read as a cheap shortcut. Marey turns compliance and licensed data into a production argument. Belief Studio brings the format to publishable form through a precise brief, human taste and the ability to see the work through to the end.
The measurable effect of AI in video production today lies in lowering the entry barrier for projects that would never have been made under a classic production budget. The automatic creation of an “AI film” does not yet belong among these results.
Comments