Trained to Imagine
Rendering
Creative Intelligence
AI Production / Hollywood / Legal

Netflix used generative AI on roughly 300 titles this year

Source: Variety ·

Netflix disclosed in its Q2 earnings that around 300 titles across its library used generative AI somewhere in production this year — spanning concept and previsualization through post-production and release, rather than concentrated in any single stage.

TTI Takeaway

The number that matters is not 300, it is the spread across the pipeline. Adoption is not arriving as AI-generated shows; it is arriving as AI inside conventional production at every stage. That is exactly the shape of demand a hybrid pipeline serves, and it is why buyers now ask vendors how AI fits their existing process rather than whether it does.

Analysis

Earnings letters are written for shareholders, and that is what gives Netflix’s figure its weight. Around 300 titles using generative AI in a single year is not presented as an experiment or a showcase; it is presented as an operating efficiency, in the same document that reports revenue.

According to Variety, the second-quarter shareholder letter said the technology is used at every level of production, from concept and pre-visualisation through post-production and release. Netflix singled out the Indian sports thriller “Glory,” the Brazilian football miniseries “Brasil 70: A Saga do Tri” and the docuseries “The American Experiment,” where it said the tools helped create “highly complex sequences” including enhanced crowd sizes and battle scenes. The letter said Netflix is using these tools “to deliver higher-quality output more quickly and at a lower cost than traditional methods,” and that in some cases productions would otherwise have had to leave out key shots and sequences. On the earnings call, co-CEO Ted Sarandos pointed to 17 minutes of AI-enhanced footage in “The American Experiment,” which he said was produced “twice as fast and at half the cost of previous options.”

The examples are telling. Crowd extension and battle sequences are long-standing VFX problems, traditionally handled with crowd simulation, tiling and digital doubles, and here they sit inside conventional shows made by conventional crews. The claim is not that AI made a series. It is that AI made a scale of scene affordable within a budget that would otherwise have cut it. That is the most realistic adoption path for scripted and documentary work alike, and it is where vendors will be judged on how well they integrate with existing editorial and finishing, not on standalone generation.

Sarandos framed it explicitly as a tool for creatives, saying that “movies are being made by people who make movies,” and Variety recalls Netflix’s March acquisition of InterPositive, the company founded by Ben Affleck, which Sarandos described as helping filmmakers improve their films using their own dailies and production materials. That detail matters for rights. Working from a production’s own footage is a much cleaner provenance position than generating from a general-purpose model, and it suggests where studios will want to draw the line when they write vendor terms.

What the disclosure leaves out matters just as much. “Roughly 300 titles” counts programmes that touched the tools somewhere; it is not a measure of how much of each was generated, and the letter as reported does not break down which stages account for most of the use. Nor does it say how performer consent or guild terms, including the new SAG-AFTRA provisions on synthetic performers, applied in these cases. Our view is that the headline number will be quoted far more often than it is examined, and that the figure with more practical consequence is the cost claim. Twice as fast at half the cost, attached to one specific 17-minute block, is the kind of benchmark that will surface in budget conversations with streamers from here on, whether or not a given show’s material resembles it.