How a Machine Ended Up in the Vertical Drama Business, Three Ways
People who work in vertical drama give three different answers about artificial intelligence in 2026, depending on which end of the business they sit in. The first faction sees a cost line inside an established app, applied quietly Accidental Surrogate for Alpha online to a portion of the slate. A second group has produced complete series that no camera ever touched and dropped them directly into social streams. A third faction spends its time bickering about what the first two are doing to the trade. In a genre already engineered for pace, all three are happening at the same time. Short dramas are vertical series of 60 to 100 second episodes, usually 50 to 100 of them, and Western producers have noted budgets of roughly $150,000 to $300,000 for a shoot of eight to ten days. Any automated production pipeline gets judged against that figure.
Inside the Apps: AI as a Line Item
The version of this story that matters in commercial terms is also the least conspicuous. Within the leading apps, generative tools get used the way any other cost-saving tech gets used, on part of the catalogue rather than all of it, and rarely with a banner announcing it. The economics explain the enthusiasm. Chinese production company Jiuzhou Culture told state media that a 70-episode series cost $100,000 to $150,000 to make. Set against numbers like those, any process that removes a shooting week is worth piloting. The prize is big enough to justify the test run, since Sensor Tower puts 2025 in-app revenue across short drama apps at about $2.8 billion to $3 billion, more than double the year before. Two current examples give the spectrum of what the apps are actually putting out.
The Eighty Percent Slate at DramaWave
DramaWave is the clearest case of a company treating generated output as the production line rather than a decoration. The app is published by SKYWORK AI PTE LTD, part of Kunlun Tech’s Skywork business, and it went live in September 2024. According to the outlet 36Kr, by the second three-month period of 2026, more than 80 percent of DramaWave’s newly released content was AI-generated and its annualized revenue had topped $700 million. An sector report, for scale, put that annualized figure at roughly $120 million in March 2025, with more than 10 million monthly users. About 13 percent of the microdrama market outside China belongs to DramaWave, by Media Partners Asia estimates. The listing claims more than 30,000 dramas in 18 languages, while Google Play shows 100 million-plus installs. A weekly pass runs about $7.99 to $19.99 in the US App Store as of September 2026, with a monthly plan near $36.99, and those rates vary by market and change without much notice.
The Animated Motion Comic at ReelShort
ReelShort went more limited, setting animated titles alongside its live-action catalog. Shattered Vows, a 2026 ReelShort series running 52 episodes, gets described as an AI-generated animated motion comic. Heiress Vanessa Windsor, who learns that her husband and her best friend helped her sister’s killer escape, carries the narrative, and the page lists no cast at all, which is the practical tell for this kind of production. There is a second ReelShort title as well, The Lost Heir, which runs 58 episodes and is equally an animated motion comic rather than live action. The lead is Arthur Lancaster, who loses his memory saving his brother, grows up on a farm under the name Johnny, and eventually exposes the foster sister who spent years framing him. Anyone searching should know that a separate ShortMax series also called Shattered Vows is an entirely different show with a different storyline. Both run at full series length, so neither ReelShort title is a token experiment.
Outside the Apps, When the Whole Show Is Generated
The second strand skipped the app stores entirely. Fruit Love Island is an AI-made vertical micro-drama parodying the reality dating show Love Island, in which anthropomorphic fruit contestants move into a villa, couple up and recouple, complete with a spoof of the Casa Amor twist. Visuals, voice-overs and scripts all came from generative tools, episodes run about two to four minutes, and Forbes likened the look to a children’s CGI-animated film. Voting on contestant choices happened through Google Forms, a participatory hook none of the commercial apps has equalled. It did not come out of nowhere, since a wider surge of ai fruits clips had been going around since February 2026, tracked by Know Your Meme to a TikTok account that uploaded a strawberry’s affair with her eggplant boss. The growth curve disquieted people in the sector, since episode one went up on March 13 or 14, 2026, and by March 22 the account had passed 3 million followers and 300 million views. By April 6, TODAY noted total views of 214 million, with episodes averaging more than 10 million apiece. That show has its own full write-up elsewhere in this set of articles, so the point worth keeping here is more limited: one anonymous account reached app-scale attention with no camera, no cast and no budget anybody can name.
Reading the Backlash Closely
Criticism came almost as fast as the views, and critics commonly called the series AI slop. That phrase is doing several jobs at once, and it helps to pull them apart. The aesthetic grievance comes first, aimed at work that is low-cost to produce and frictionless to swipe past. The second complaint is about credit, and comedy creator and voice actor Joy Ofodu told TODAY that the series was likely inspired by her own talking-fruit videos, which points at the awkward fact that generative tools can copy a style without ever naming the person who invented it. The documented record itself is the third objection. The original Ai Cinema account has an anonymous and unconfirmed creator, several online names have professed authorship without proof, no website has been verified as the show’s official home although several present themselves that way, and copycat accounts began posting imitation episodes within weeks. The chronology itself has soft edges, given that the date of the Season 2 first episode is taken from the video rather than published by anyone. Reactions were not uniformly negative either, since Love Island USA’s JaNa Craig and Kaylor Martin interacted with the series and remade clips with human actors, while Season 7 winner Amaya Espinal said she would never watch it.
Where AI Shows Up and What Is Actually Verified
Most of the work in covering this subject is sorting the confirmed from the assumed. Below, the table splits the places generated content turns up from the claims tied to them. People still type ai fruit love island into search boxes months after the villa closed, which tells you how much of the discussion is fuelled by one title rather than by the apps discreetly releasing generated episodes every week. Only what a named source actually proved is listed in each entry.
| Where AI turns up | Example | What is actually verified |
|---|---|---|
| Most of an app’s new release slate | DramaWave | Per 36Kr, over 80 percent of newly released content AI-generated, with annualized revenue past $700 million by the second quarter of 2026 |
| Animated titles within a live-action library | ReelShort’s Shattered Vows, 52 episodes | Described as an AI-generated animated motion comic, with no cast listed |
| Animation without a stated AI attribution | ReelShort’s The Lost Heir, 58 episodes | Flagged as an animated motion comic rather than live action |
| A whole series made outside the apps | Fruit Love Island | AI used for visuals, voice-overs and scripts, with episodes of about two to four minutes |
| Authorship behind a viral AI series | The Ai Cinema account | Anonymous and unconfirmed, with several unproven claims of authorship |
| Sites claiming to be the official home | Fruit Love Island fan and search-optimized sites | None verified as official, and imitation accounts exist |
| Release dates for generated material | Fruit Love Island Season 2, episode one | Date taken from the video itself, never released by the creator |
Whether generated vertical drama is good is not settled by any of this, and that argument will outlive the fruit. The public record does show this much: the form’s economics rewarded automation before viewers asked for it, and the apps moved first. For viewers the practical guidance stays dull: check the account handle before you rely on an upload, take fan sites and inferred dates as tentative, and never download an unofficial app or a modified build advertising free episodes or unlimited coins. Having spent 2026 proving it can produce attention at scale, the genre leaves an open question for 2027: whether anyone can tell the difference on a phone screen.
