AI micro-dramas: the vertical format explained

Vertical, 60 seconds an episode, cliffhanger every time. Micro-dramas are the fastest-growing story format in short-form, and AI just removed the crew.

Published 2026-07-30 by Cat Claw, the UK's create-and-distribute AI studio. About a 8 minute read.

The short version

A micro-drama is a scripted vertical series told in roughly 60-second episodes, each ending on a cliffhanger.
The format works because a full story beat fits inside the attention window short-form already trained people to have.
AI production means one person can script, cast, shoot and assemble an episode without a crew or a location.
The hard parts are character consistency between shots and continuity between scenes, and both are solvable.
Story still decides everything. The tools remove the crew, not the writing.

Micro-dramas are scripted stories told in vertical episodes of roughly a minute, each ending on a hook that makes the next one unskippable. The format grew up on Chinese short-video apps, moved to TikTok and Reels, and now supports an entire industry of dedicated apps.

Until recently making one meant what making any drama means: actors, locations, a camera operator, days of shooting for minutes of screen time. AI changes what a production team has to be. This guide covers what the format actually is, why it works on the brain the way it does, and how an episode gets made now.

What is a micro-drama?

Quick answer. A micro-drama is a scripted, serialised story told in vertical episodes of about 60 seconds, each built around one dramatic beat and ending on a cliffhanger. An AI micro-drama is one produced with generative tools: the script, the cast, the shots and the edit are all generated rather than filmed, so a single creator can make what used to need a production crew.

The structure is closer to a soap opera than a short film. Each episode does one thing: a revelation, a betrayal, a confrontation, a reunion. There's no room for a second act, so every scene carries plot.

Genres cluster hard. Noir, romance, revenge, secret-identity and workplace drama dominate, because they generate cliffhangers naturally. If your premise doesn't produce an obvious "and then what" every 60 seconds, it's a short film, not a micro-drama.

Why the format works

Short-form video spent a decade training an enormous audience to give a video about 3 seconds to justify itself and then to keep going if it does. Micro-dramas are the first scripted format built for that behaviour rather than fighting it.

A whole story beat fits in the window. You get a complete satisfying unit, not a fragment of something longer.
The cliffhanger does the work an algorithm usually has to. Viewers choose the next episode rather than being served it.
Vertical framing suits faces, and micro-drama is almost entirely faces reacting to each other.
Serialisation compounds. Every episode is an advert for the series, and a viewer who reaches episode 3 usually finishes.

For creators the appeal is that it rewards writing over production budget. A well-constructed betrayal lands whether it cost £50 or £50,000.

How an AI micro-drama gets made

In Cat Claw's Drama Studio a show moves through 5 stages, and you can hop back to any of them before publishing.

Drafting. You start from a premise or a template. The studio ships pre-built shows across genres like noir and romantic comedy, each arriving with a cast and a prop shelf already defined, which is the fastest way to learn the shape of the format.
Casting. Characters are locked before shooting. Each gets a role type: Lead, Ally, Antagonist or Supporting. Locking the cast is what keeps a face the same person from scene to scene.
Shooting. Every scene is shot as one or more takes. You generate, watch, and generate again until one is right, then approve exactly one take per scene. The approved takes in order are the episode.
Assembling. The approved takes render into a single episode file in the background while you carry on working.
Publishing. The finished episode gets a share link, and goes out to social from the same studio.

The approve-one-take-per-scene rule is the quiet good idea here. It stops a show becoming an unmanageable pile of variations, and it means the episode always exists in a coherent state rather than needing a final assembly decision at the end.

The two genuinely hard parts

Anyone who has tried to make a story with generative video has hit both of these within an hour.

Keeping a character the same person

Generate the same character twice from the same description and you'll usually get 2 different people. For a drama that's fatal: the audience needs to recognise a face across a cut without thinking about it.

The fix is locking characters at the casting stage rather than describing them again per shot. Cast once, and every scene draws from the same locked identity. If a character needs to be a specific real person, a saved Muse holds that likeness across every render.

Making one scene follow another

The other problem is spatial. Two consecutive shots in the same room should look like the same room, but a fresh generation invents a new one every time.

Drama Studio handles this by chaining a scene off the previous scene's approved last frame, and it's a per-scene choice rather than a global switch, because the right answer changes shot to shot. Auto chains when the location matches the previous scene. Linked always chains. Fresh never does, which is what you want at a location cut or a time jump.

Getting this wrong in either direction is obvious to viewers. Chain across a location cut and the new place is contaminated by the old one. Don't chain within a conversation and the room quietly changes behind your characters.

Writing a micro-drama that holds attention

Open in the middle. No setup, no establishing shots. The first line should be the second-most interesting thing that has happened today.
One beat per episode. A revelation, a betrayal, a confrontation. If you're covering 2, you have 2 episodes.
End on a question, not a resolution. The cliffhanger is the whole retention mechanism, so write the ending first and work backwards.
Keep the cast small. 3 to 4 characters is plenty, and every extra face is another identity to hold consistent.
Write for faces. The format is close-ups and reactions, so put the drama in what people do to each other in a room, not in scale or spectacle.
Say things out loud. There's no time for subtext to land. Micro-drama dialogue is direct in a way that would feel broad in a feature, and that's correct for the format.

Who should be making these

Writers, most obviously. It's the first video format where writing ability translates more directly into results than production budget, which inverts the usual barrier.

Beyond that: creators who want a reason for an audience to come back rather than to scroll past one good video, and brands willing to tell a story rather than run an ad. Serialised fiction with a product woven through it holds attention that a 30-second spot cannot buy.

It's also, bluntly, the cheapest way to find out whether you can write. A premise either produces a cliffhanger every 60 seconds or it doesn't, and you'll know inside one episode.

Frequently asked questions

What is a micro-drama?

A micro-drama is a scripted, serialised story told in vertical episodes of roughly 60 seconds, each built around a single dramatic beat and ending on a cliffhanger. The format grew out of Chinese short-video platforms and now runs across TikTok, Reels, YouTube Shorts and dedicated micro-drama apps.

How long is a micro-drama episode?

Usually around 60 seconds, sometimes as short as 30. The constraint is the point: one beat per episode, ending on a hook. A series typically runs anywhere from 10 to 100 episodes.

How is an AI micro-drama different from a normal one?

Only in how it's produced. The script, cast, shots and edit are generated rather than filmed, so there's no crew, no location and no shooting schedule. The storytelling requirements are identical, and the writing still decides whether it works.

Can one person make a micro-drama?

Yes, and that's the main change. Drama Studio takes a show from script through casting, shooting and assembly to publishing without anyone else involved. What used to need a crew and days of shooting is now one person and an afternoon.

How do you keep characters looking the same across scenes?

By locking the cast before shooting rather than re-describing characters shot by shot. In Drama Studio, characters are cast once and every scene draws from that locked identity. For a specific real person, a saved Muse keeps that likeness consistent across every render.

How do you keep the setting consistent between scenes?

By chaining a scene off the previous scene's approved last frame. Drama Studio makes this a per-scene choice: auto chains when the location matches the previous scene, linked always chains, and fresh never does. That last option is what you want at a location cut or time jump, where chaining would drag the old setting into a new place.

What genres work best for micro-dramas?

Anything that generates cliffhangers naturally: noir, romance, revenge, secret identities, workplace and family drama. If your premise doesn't produce an obvious "and then what" every 60 seconds, it's better suited to a short film.

Do I need to write the script myself?

You can, or you can start from a template. Drama Studio ships pre-built shows across genres like noir and romantic comedy, each with a cast and props already defined. Starting from a template is the quickest way to learn the format's rhythm before writing your own.

Where do you publish AI micro-dramas?

Vertical short-form platforms: TikTok, Instagram Reels, YouTube Shorts, plus dedicated micro-drama apps. Cat Claw gives each finished episode a share link and posts to around 15 social platforms from the same studio.

Can brands use micro-dramas?

Yes, and serialised branded fiction holds attention a 30-second spot can't. The rule is that the story has to work on its own terms. A micro-drama that exists to sell something reads as an ad by episode 2, and viewers stop choosing to continue.

How many episodes should a series be?

Start with 5 to 10 and see whether people finish. Reaching episode 3 is the signal that matters most: viewers who get that far usually finish a series. Build out from a premise that's proven it can sustain hooks rather than committing to 50 episodes up front.

Is it expensive to make an AI micro-drama?

Far cheaper than filmed drama, though not free. Each take costs credits and you'll generate several before approving one, so the real cost sits in iteration rather than in crew or locations. Writing tightly, with a small cast and few locations, keeps that cost down as well as making a better episode.

Direct your own micro-drama.

Script it, cast it, shoot it and publish it from one studio. No crew, no location, no shooting schedule. Free to sign up, no card needed.

What micro-dramas are, why the vertical short-form format exploded, and how AI lets one person write, cast, shoot and publish an episode without a crew.