Every major transformation in cinema - the introduction of sound, the arrival of colour, the shift to digital - was met with the same combination of excitement and alarm. AI is no different, except the pace is unlike anything the industry has seen before. What took decades previously is happening in years. What took years is happening in months.
This is a factual account of how that transformation unfolded - what actually happened, year by year, from the first mainstream applications of AI in Hollywood to the emergence of a fully new category of filmmaking.
Year by year: the timeline
AI enters Hollywood through the back door
The first significant uses of AI in mainstream film were invisible to most viewers - and that was the point. Industrial Light & Magic (ILM) used machine learning-based neural networks to de-age and voice-synthesise Mark Hamill as a young Luke Skywalker in The Book of Boba Fett, producing results that previous CGI techniques could not match at any budget.
Films like The Mandalorian were pioneering LED volume stages - virtual production environments where AI-assisted real-time rendering replaced location shoots entirely. The technology was expensive, proprietary, and invisible. Almost no one outside the industry knew it existed.
Simultaneously, deepfake technology - initially a fringe internet phenomenon - was being studied seriously by studios for its implications on actor likeness rights. SAG-AFTRA began internal discussions about digital replication that would define their 2023 strike demands three years later.
Diffusion models change concept art overnight
In August 2022, Stability AI released Stable Diffusion publicly. In the same period, Midjourney launched. Within weeks, the tools were generating photorealistic images from text prompts with a quality that stunned the industry.
For film pre-production, the immediate impact was on concept art and storyboarding. Designers who previously spent days on a single piece of concept art were iterating through dozens of variations in hours. The tools did not replace concept artists - they changed what concept artists could do in a day.
Generative AI companies received $1.37 billion in investment in 2022 - more than the combined total of the previous five years. The financial signal was unmistakable. Runway launched its first AI Film Festival, with a panel including filmmaker Darren Aronofsky discussing the future of AI in cinema.
The strike that made AI a labour question
On May 2, 2023, the Writers Guild of America went on strike. 97.9% of WGA members voted in favour. For the first time in its history, AI was a primary negotiating issue - not a background concern, but a frontline demand.
The strike lasted 148 days. Joined by SAG-AFTRA, it was the first simultaneous writers and actors strike in over 60 years. The economic impact on California alone exceeded $3 billion. Productions including Stranger Things halted. Late-night television went dark.
The settlement reached on September 27, 2023 established landmark rules: AI cannot be considered a “writer” on a project. AI-generated material cannot be used as source material. Studios must disclose to writers when they are given AI-generated content to work with. Writers cannot be forced to use AI tools.
Meanwhile, the first AI video generation tools aimed at creators went public. Pika Labs launched on Discord and rapidly built a substantial user base. Runway released Gen-2. The tools were limited - short clips, inconsistent characters, obvious artefacts - but the trajectory was visible to anyone paying attention.
Sora's demo, Runway's partnership, and the first AI series
On February 15, 2024, OpenAI released demonstration clips from Sora - a model generating up to one minute of photorealistic video from text. A Tokyo street scene. A fox bounding through snow. Realistic ocean waves. The internet collectively stopped. Hollywood executives began making calls the same day.
Runway secured a partnership with Lionsgate Studios to train its AI model on Lionsgate's film catalogue - the first formal agreement between a major studio and an AI video company. In December 2024, Runway launched a talent network connecting independent AI creators to studios and production companies.
Sora launched to ChatGPT Plus and Pro subscribers in December 2024, after months of selective access for filmmakers and researchers. Expectations were very high. The results were impressive but inconsistent - a gap between the curated demo and real-world outputs that would define Sora's difficult public life.
AI-generated music adoption in film and video production grew from 12.5% in 2023 to over 50% by the end of 2024, according to MIT AI Film Hack data. The sound design barrier - previously one of the last holdouts of AI in production - had fallen.
Native audio, complete pipelines, and Netflix's first AI show
In May 2025, Google DeepMind released Veo 3 - the first major AI video model to natively generate video and audio in a single pipeline. No post-processing. No separate sound design layer. One prompt, one complete output with synchronised audio, dialogue, and ambient sound.
Netflix used generative AI in production for the first time in The Eternaut, a major Spanish-language series. Netflix co-CEO Ted Sarandos called it an “incredible opportunity.” The show premiered to significant viewership - and generated significant debate about disclosure and labour.
Runway's AI Film Festival moved to Lincoln Center's Alice Tully Hall in New York on June 5, 2025. A line formed down Broadway before doors opened. The prize pool reached over $135,000, including a Grand Prix of $20,000 and one million Runway credits. The festival's scope had expanded beyond film to gaming, fashion, design, and advertising.
Turkish TV powerhouse Ay Yapim announced Castle Walls - three 15-minute episodes set in Iran's Alamut Castle, written by television veterans and produced with AI generation throughout. It was set to premiere simultaneously on Turkish free-TV and a global streaming platform.
Amazon invested in Fable, whose Showrunner platform lets users create scenes and entire episodes of AI-generated TV shows - described by Variety as “the Netflix of AI.”
Sora shuts down. The market consolidates.
In late March 2026, OpenAI quietly discontinued Sora as a consumer product - less than six months after its public launch. The reasons were primarily economic: video generation at scale is computationally expensive, and Sora's usage had not justified its infrastructure costs. OpenAI redirected resources toward reasoning models, coding tools, and agentic workflows.
The exit reshuffled the market. Runway, having matured from a promising indie tool into a platform used by major studios, solidified its position as the professional standard. Kling AI - built by Chinese company Kuaishou - had become the go-to for long-form clips and accessible pricing. Google's Veo, integrated into the YouTube ecosystem, became the dominant enterprise option.
The lesson Sora's arc taught the industry: a stunning demo is not a product. Consistency at scale is the hard problem. And the companies that had been iterating steadily - rather than arriving with a single extraordinary launch - were the ones still standing.
The festivals that did not exist five years ago
In 2020, there were no film festivals dedicated to AI cinema. By 2026, there are dozens - across six continents, with prizes that rival established independent film competitions.
These festivals have created something that did not exist before: a professional context for AI filmmakers. Career paths. Industry contacts. Distribution opportunities. A community. The infrastructure of a legitimate new genre of cinema.
The first AI-generated series
The most significant indicator of AI's arrival in mainstream production is not a festival or a tool - it is the commissioning of series by broadcasters and streaming platforms.
The question is no longer whether AI will be used in series production. It is about how much, disclosed how, and governed by what rules.
What Hollywood actually thinks
Hollywood's relationship with AI is not a simple story of resistance or embrace. It is more complicated - and more interesting.
The major studios are simultaneously funding AI research, negotiating labour agreements that restrict AI use, and exploring AI for production efficiency. Warner Bros. signed deals with AI-driven management systems years before generative AI became mainstream. Disney has reportedly explored content partnerships with AI companies. The same executives who publicly caution about AI are privately asking their production teams to find ways to use it.
The WGA agreement - for all its historic significance - expires in May 2026. The next round of negotiations will take place in a landscape where AI tools are demonstrably more capable than when the 2023 contract was written. What the guild secured in 2023 was a beachhead, not a permanent settlement.
Darren Hendler, known for his VFX work on Avengers: Endgame, put it well: Hollywood's automated future is not one that removes humans from the frame. It is one where creative people remain creative - with more powerful tools in their hands.
What it means for independent filmmakers
The story of AI and Hollywood - strikes, studio deals, streaming partnerships - is not the most important story here. The most important story is what's happening at the edges: the independent filmmaker in Istanbul who made a pilot without a crew. The YouTuber in São Paulo producing a serialised sci-fi series from a laptop. The screenwriter in Seoul who can now show a producer what their script looks like moving, rather than describing it on paper.
AI has done something that decades of “democratising filmmaking” rhetoric could not: it has made the barrier to entry genuinely low. Not low enough that anyone can make a great film - taste, vision, and storytelling instinct still matter enormously - but low enough that budget is no longer the primary constraint.
In 2020, a pilot episode cost between $100,000 and $500,000 to produce at professional quality. In 2025, the same creative result - imperfect in some ways, surprising in others - costs between $100 and $1,000. That is not an incremental change. That is a structural transformation of who gets to tell stories.
What comes next
The next phase of AI in filmmaking will be defined by three things:
Character consistency. The persistent technical limitation - keeping a character visually identical across scenes without manual intervention - is the primary unsolved problem for long-form AI production. Every major lab has it on their roadmap. When it is solved, episodic AI production becomes genuinely scalable.
Labour and disclosure. The WGA contract expires in 2026. SAG-AFTRA's agreement on digital likeness rights is already being tested by productions in markets without equivalent protections. The legal and ethical framework for AI in production is still being written - and the outcome will shape how the tools are used for decades.
New formats native to AI. The most interesting development is not AI making traditional films cheaper. It is AI enabling formats that could not exist before: personalised narrative content, interactive cinema, procedurally generated worlds that respond to viewers in real time. These are not improvements on existing formats. They are new ones.
We are, by any honest measure, at the beginning of this. The tools that will define AI filmmaking in 2030 do not exist yet. The creators who will use them are only now learning what's possible.
Be part of what comes next
Framrlab is built for filmmakers who do not want to wait for the industry to catch up. From screenplay to final cut - in one pipeline.