Welcome to the third edition of the Starring Nobody Intelligence Report, the monthly deep-dive tracking the data, deals, and fault lines as Hollywood navigates the most disruptive shift in its hundred-year history.

Our inaugural report ended on a prediction: 2026 would see the first AI-generated feature film reach commercial distribution. It happened in April. It just didn't happen here. While Los Angeles spent the month at the bargaining table, the action moved east, to a Chinese studio's box office, to a leaderboard topped by Chinese models, and to a courtroom where the record industry is quietly choosing sides. April was the month AI Hollywood stopped being a Hollywood story.

On April 28, Bona Film Group won regulatory approval to release a 90-minute AI-generated sci-fi feature in Chinese cinemas, built with ByteDance's video tools and classified, tellingly, as "animation." This is the milestone our first edition forecast for 2026, the first AI-primary feature with real theatrical distribution. Every major Hollywood studio has spent a year announcing AI ambitions and convening committees. A Chinese studio simply shipped one to the box office. The first AI feature on a commercial screen did not come from Burbank, and that fact reorders everything that follows.

Last month we covered the post-Sora tool race as a three-way fight among Kling, Veo, and Runway. April redrew that map. Alibaba's HappyHorse-1.0 launched straight into the top tier of the industry's main quality benchmark, and ByteDance's Seedance 2.0 sits beside it. The two best AI video models in the world are now both Chinese. The American product that defined the category is gone, and the tools filling the vacuum are being built an ocean away, under a very different set of rules about training data and consent.

Music has lived in the back half of this report for two editions. In April it earned the front. Warner Music settled its lawsuits against Suno and Udio and signed licensing deals with both, following Universal's late-2025 settlement. That leaves Sony Music as the last major label still litigating, a lone holdout heading into a fair-use ruling expected this summer that could reset the price of every AI song ever generated. The labels have split into two camps: those monetizing the machines and those still trying to sue them. How that split resolves will decide what AI music costs for the next decade.

Stories we have tracked in prior editions that advanced in April, but did not reinvent themselves:

The WGA, whose 2023 AI provisions we broke down in our first edition, ratified a four-year deal with the AMPTP: a record $321 million health-plan infusion and notification rights if studios license scripts for AI training, but no payment for the training itself. A defensible peace, and the template SAG-AFTRA now negotiates against. We will track the actors' response below.

The shutdown we led with last month became final on April 26, when OpenAI pulled the Sora web and app for good (the API limps to September). The product is buried. The market it created, as Story #2 shows, is very much alive.

Formal talks with the AMPTP resumed in April against the June 30 expiry, with the "Tilly tax" we introduced last month as the marquee demand. The real fight is next month.

When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning — constraints, edge cases, examples, tone — and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.

89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.

Start flowing free

The key metrics forming our baseline, updated with April's developments. Changes are flagged.

For two editions our Deal Room tracked Big Tech buying Hollywood. In April the action moved to music, where the deals and the lawsuits are now happening at the same table.

Warner Music Group settled its copyright suits against Suno and Udio and signed licensing arrangements with both, joining Universal, which settled and struck its own deal in late 2025. The terms reportedly pair compensation for past training with go-forward licenses for AI music platforms. Sony Music did not follow. It remains the only major still actively litigating both companies, even as a multibillion-dollar industry suit over 20,000-plus songs grinds forward and a pivotal fair-use ruling looms this summer.

This is the first time the major labels have visibly disagreed on AI strategy, and the disagreement is a bet on one court decision. If the summer ruling lands on the side of fair use, the licenses Universal and Warner signed lose leverage and Sony's holdout looks shrewd. If it lands the other way, the settlers locked in revenue early and Sony is exposed. Either way, the people who actually made the training data, the session musicians we wrote about in our first edition, are not party to any of these deals.

With the writers settled, the actors resumed formal talks with the AMPTP in April against a hard June 30 expiry. The marquee demand remains the "Tilly tax" we introduced last month: a penalty requiring studios to pay a fee for using a synthetic performer in place of a human. The question for May is whether SAG-AFTRA can hold a line on synthetic-performer economics that the writers just walked back on training data. The actors have one thing the writers did not, a public face for the threat. It is hard to rally a strike around datasets. It is easy to rally one around a synthetic actress with no agent and no off switch.

The story is no longer which Western tool wins. It is that the Western tool lost.

The top two models on the industry's main quality benchmark, both Chinese. Seedance is a unified audio-video model with phoneme-level lip-sync across multiple languages, the first tool to meaningfully collapse the video-plus-sound pipeline into a single pass. It is also the engine behind the Bona theatrical feature. HappyHorse arrived in April and went straight to the front.

Veo remains the enterprise all-rounder a US studio would build a pipeline on, the safe institutional choice. Kling stays the value play at roughly ten cents a second. But the momentum, the leaderboard, and the first theatrical credit all now point east. The center of gravity for AI video production has moved, and it is not coming back on its own.

Above the label settlements sits a quieter front: independent artists and small labels have filed their own class actions over recordings used in training without consent. They are not getting the Warner treatment. The major-label deals carve up a market the independents helped build the technology on, and the summer fair-use ruling will hit them hardest of all, with the least leverage to absorb it.

The Bona feature is the VFX story this month. A theatrical-length film built with a fraction of the crew a traditional animated feature requires is the productivity-multiplier pitch made literal. The work got made. It got distributed. It simply employed almost none of the artists who, as we wrote in our first edition, spent years generating the very training data that learned to replace them.

The throughline of this report has been a paradox: AI is gutting the middle class of entertainment work while handing the tools of cinema to anyone with a laptop. April advanced the second half of that paradox past the point of argument. The first AI-generated feature to reach a commercial theater was approved in China, made with Chinese tools, by a studio that did not wait for Hollywood's permission, unions, or consensus. The democratization we have tracked from contest submissions has reached the box office. The frameworks being negotiated in Los Angeles are being built to govern an industry that is no longer centered there.

[Featured video slot]

Drop this edition's standout AI short or clip here.

By spring 2027, the strike scenario that defined the 2023 cycle will have faded, and the action will have moved from the bargaining table to the box office and the courtroom. The defining question shifts from whether AI can be used to who is liable when it is, and where the work happens. April suggests the answer to that last part is increasingly "not here." Expect a US studio to greenlight an AI-primary feature within the year, if only to avoid ceding the form entirely to Beijing.

The bifurcation thesis we laid out in our first edition (prestige content stays human, the streaming ocean goes synthetic) held in April, with one revision: the synthetic ocean may not be American. China shipping the first theatrical AI feature and building the best tools raises a scenario we underweighted, that the low-cost, high-volume tier of global filmed content gets produced offshore with offshore tools, while Hollywood's careful contracts govern a shrinking, premium island. The contracts assume Hollywood is still the center. April is the first month that assumption looked shaky.