How Micro-Drama Scripts Work: The 90-Second Beat That Sells the Next Episode
Micro-dramas are paid for one episode at a time. Here is the beat sheet inside a 90-second episode, the economics behind it, and the mistakes that end a series early.
- #Micro Drama
- #Screenwriting
- #Vertical Video
- #Story Structure
- #ReelShort
430,000 titles in eight months
China's broadcasting regulator counted 430,000 micro-dramas released between January and August 2026, thirteen times the output of all of 2025. More than 90% were made with AI, and the country now has more than 800 million micro-drama viewers. (Xinhua)
That number gets read as a production story. It is an attention story.
When 430,000 titles can be produced in eight months, nobody wins by being produced. The competition is for the first three seconds and for the next tap. A script that ignores this gets scrolled past.
The money explains why the competition is so sharp. Outside China, micro-drama apps earned $1.52 billion between January and August 2025, up 194.9% year on year, on roughly 730 million downloads, up 370.4% (Xinhua). A 2026 industry forecast puts the overseas market above $6 billion, 67% higher than 2025 (China Daily). ReelShort is on track for $1.05 billion in revenue this year and its first profit, with an estimated 29% share of the category against DramaBox's estimated 21% (Variety, citing MPA).
The concentration is even sharper than the growth. In the first quarter of 2026, ReelShort and DramaBox each booked an estimated $140 million in in-app revenue (Sensor Tower). The rest of the market is fighting for what is left.
In China, the domestic market reached 50.5 billion yuan in 2024, passing the country's cinema box office for the first time (Caixin).
The format, in numbers
Micro-drama is a loose label. In practice, the format that ReelShort and DramaBox sell has a narrow shape:
| Element | Typical range |
|---|---|
| Episode length | 60–120 seconds, usually built around 90 |
| Episodes per season | 60–100 |
| Orientation | 9:16, designed for one hand |
| Free episodes before the paywall | 5–15; moves with promotions |
| Season runtime | 1–3 hours |
The 60-to-100-episode season is a business shape, not an artistic one. Episodes are sold one at a time — coins on ReelShort, a similar currency elsewhere — so a season that runs 90 minutes can be priced in dozens of small decisions instead of one subscription. The US Commercial Service's country guide describes episodes from 30 seconds to five minutes and seasons of 60 to 100 episodes (trade.gov). ReelShort's free opening run is usually reported at 5 to 13 episodes, depending on which of its own guides you read, and one catalogue measurement across 86 series found a median of 10 free episodes (DramaCatalog, Reel Shorts, an unofficial fan-maintained site despite the name).
That free run is the whole business. The viewer is not deciding whether to start a series. They are deciding, at the end of the free run, whether to spend money to find out what happens next.
The paywall is a story beat
This is where most writing advice about "short-form" goes wrong. It treats the episode as a small film with a beginning, middle, and end. A micro-drama episode is not a film. It is a free sample with a price attached to the next one.
The structure follows from that. Every episode has to leave a debt unpaid — a question, a threat, a promise broken — that the viewer can only settle by tapping. The season arc matters, but it is not what gets sold at the paywall. What gets sold is the next 90 seconds.
The audience data is brutal about how little room there is. In 2024, 45% of Chinese micro-dramas on short-video platforms had fewer than one million plays, and only 4.3% passed 100 million (2024 Micro-Short Drama Industry White Paper). Most titles never get sampled at all. A slow first episode is not a slow start; it is the end.
The 90-second beat sheet
The scripts that survive the paywall tend to hit the same six beats. The timings below are for a 90-second episode; a 60-second cut has to merge the escalation and the turn.
| Time | Beat | Job |
|---|---|---|
| 0–3s | Hook | Disrupt something. No context, no introduction. |
| 3–15s | Orientation | Who is in the scene, what they want, what it costs. Two lines, maximum. |
| 15–45s | Escalation | One obstacle, one tactic, one refusal. |
| 45–70s | Turn | New information that changes what the scene means. |
| 70–85s | Cost | The consequence lands. Someone loses something on screen. |
| 85–90s | Cut | A new question, not an answer. |
Treat the timings as a working template, not a rule. The shows that break it usually do so by making the hook do two jobs at once: it disrupts the scene and plants the season question.
The word budget is arithmetic, not taste. Normal dramatic speech runs at two and a half to three words per second. Ninety seconds is about 200 words of dialogue once you leave room for silence, reaction shots, and action. That is one page, and it is why exposition dies in this format. You cannot explain the backstory and also turn the scene.
Here is the skeleton applied to an ordinary premise — a wedding planner hired by the man who left her at the altar:
- 0–3s. She opens the venue door. He is holding a contract with her signature on it. She has never seen the contract.
- 3–15s. He is the client. The wedding is in six days. She is the planner.
- 15–45s. She tries to hand the job to a colleague. He refuses, and points at the penalty clause.
- 45–70s. The bride's name on the contract is her sister's.
- 70–85s. She takes the job. The clause would bankrupt her, and her sister cannot find out.
- 85–90s. He texts someone: "She took it."
Nothing in that episode is resolved. Everything in it raises the price of stopping.
Three ways an episode loses the tap
The hook explains instead of disrupting. A line like "I'm a wedding planner, and this is the worst week of my life" tells the viewer what the show is about. It gives them no reason to stay. "That's my signature. I've never seen this contract" starts after the disruption, which is where the audience wants to be.
The episode resolves its own question. If the couple reconciles at 70 seconds, the viewer has nothing left to buy. The resolution belongs to the next episode. The current one should end on the cost of the decision, not on the comfort of the outcome.
The cliffhanger reveals instead of withholds. "The bride is my sister" is a reveal. It closes a question. A cliffhanger opens one: the sister walks in and says, "You promised you'd never tell him." Now the viewer needs episode two to understand what was promised.
These are not small notes. In a 90-second unit, one of them is the difference between a paid unlock and a swipe.
A season is a stack of debts
Season-level structure is where micro-drama borrows least from television and most from serialized fiction.
The first three episodes do a specific job: episode one establishes the lead and the premise, episode two puts the antagonist and the stakes on the table, and episode three delivers the first reversal, so the viewer learns that the series will keep turning. Teams that write the season as one 30-to-60-minute script and cut it into minute-long episodes still have to make those cuts land on a turn; the cut point is a writing decision, not an editing one (Page Six).
After the opening, many seasons are built in blocks of about ten episodes. Each block gets one reversal and one question. The last free episode before the paywall is the most important episode in the season: it should close one thread and open a larger one. An analysis of ReelShort's Agent of Shadows makes the same point — the free run ends on a completed payoff plus one big open question, which is stronger than either alone (ShorStory Current).
In practice, that means writing the paywall episode before episode one. If you do not know what the viewer is paying to find out, the episodes before the paywall are guessing.
What the job pays, and what it costs
The economics of the writing are as compressed as the runtime. One 2026 survey of vertical-drama work put the freelance rate at roughly $2,000 to $3,000 per script, flat fee, no residuals (Vertical Drama Jobs). Trade coverage of the Hollywood-facing side reported $5,000 to $7,000 per script, with several platforms keeping salaried in-house writing teams (Page Six). The spread is not a typo. It reflects different buyers: a platform staff writer, a freelance genre specialist, and a production company working to a fixed budget are not in the same market, even when they are writing the same format.
What the rates have in common is volume. A writer who can produce a clean 90-second beat sheet is worth more than one who can produce a beautiful scene, because the format consumes structure faster than it consumes prose.
Where AI helps, and where it does not
The 430,000-title figure is mostly an AI production story, and it would be dishonest to pretend the writing is exempt. Language models are useful at the mechanical layer: generating ten variations of a turn, keeping a character bible consistent across sixty episodes, checking that every episode ends on an open question, drafting a season arc before the first scene.
They are bad at the part that decides whether the series works. A model does not know which line is the hook, because the hook is not a property of the line. It is a property of what the viewer already wants at that second. That judgment comes from watching the cut, reading the platform data, and rewriting.
I built MicroDrama Script AI for the mechanical layer. You give it a genre, a main character, and the twist, and it produces a title, logline, character bible, and 90-second beat sheets for a ten-episode outline. The first three episodes are free and there is no account. It is a starting structure, not a verdict. The turn is still your call.
What to take away
The format is not short television. It is a serialized paid product, and every episode is a sample with a price on the next one. That single fact drives the 90-second runtime, the cliffhanger on every cut, the 60-to-100-episode season, and the way the free run ends.
If you are writing one, the test is simple. Read the last five seconds of the episode. If a viewer can stop there without an unanswered question, the episode is not finished. If they cannot, you have done the job the format pays for.
Sources
- Xinhua, "China backs live-action dramas as AI-made micro-dramas dominate releases," September 17, 2026.
- Xinhua, "Chinese micro-dramas race onto global stage," August 25, 2026.
- China Daily, "Southeast Asia emerges as fast-growing market for micro-dramas," November 2, 2026.
- Variety, "ReelShort on Track for $1.05 Billion Revenue, First Profit in 2026," August 6, 2026.
- Sensor Tower, "State of Short Drama Apps 2026," June 12, 2026.
- Caixin Global, "In Depth: Micro Drama Platforms Knock Down Paywalls as Viewers Lose Interest," August 22, 2025.
- US Commercial Service, "China Entertainment Micro-drama Industry," December 15, 2025.
- 2024 Micro-Short Drama Industry White Paper, summarized by Animation World Network China, May 26, 2025.
- Page Six, "Hollywood writers 'make more babysitting' than penning microdramas," June 29, 2026.
- Vertical Drama Jobs, "How Much Do Vertical Drama Writers Get Paid Per Script?" August 1, 2026.
- ShorStory Current, "Agent of Shadows, Episodes 1–10 Breakdown," October 29, 2026.