Sixty episodes, one minute each
Microdramas are one of the fastest-growing forms of entertainment, and the format exists because of production cost. What happens when AI changes the cost?
If you haven’t encountered microdramas, here’s the format. A series of fifty to a hundred episodes, each one to two minutes long, shot vertically for phones. Every episode ends on a cliffhanger. The stories are extremely direct: a humiliated wife reveals she’s a secret billionaire, a CEO falls for the cleaner, a wronged heir returns for revenge. The first ten or so episodes are free, and after that you pay per episode or subscribe. Apps like ReelShort and DramaBox have brought the format from China to the US over the last two years, and several of them are among the top-grossing entertainment apps.
In China it’s already enormous. Industry reports put the Chinese microdrama market last year at about 50 billion RMB, which is more than the country’s total box office.
I’ve been looking at the format closely, partly because I think it’s a preview of what AI video will be used for first. What strikes me is that every feature of it is shaped by cost.
A traditional TV drama costs hundreds of thousands to millions of dollars per episode. A microdrama series, all hundred episodes, is often shot in a week or two for a few hundred thousand RMB. That constraint explains the format. Vertical because it’s cheaper to shoot and fits phones. Short episodes because short scenes need fewer setups, and because a cliffhanger every minute keeps people paying. Simple locations, a mansion, an office, a hospital, a restaurant, reused constantly. Broad acting and melodramatic stories because they read instantly on a small screen and need no subtlety from a rushed shoot.
Now apply the prices I wrote about in April. A minute of raw AI-generated video costs somewhere between well under a dollar and about thirty dollars, depending on the model, before retakes. Even at a high retake rate, the generation cost for a hundred-minute series is far below the cost of actors, crew, locations and a week of shooting. Teams in China are already producing AI microdramas this year, mostly with Kling, Jimeng and similar tools.
So the first effect is obvious: a lot more microdramas, much cheaper, many made by very small teams.
The question I find more interesting is whether the format itself survives. “Micro” exists because production is expensive. If AI makes production cheap, why would a story have to be told in one-minute pieces, in vertical, in three locations? You could have more locations, bigger scenes and longer episodes for the same budget. Some of the format’s appeal is about attention and payment: the cliffhanger rhythm works on phones regardless of cost. But some of it is just what was affordable, and that part should change.
The real bottleneck right now isn’t cost. It’s consistency. A series needs the same characters across a hundred episodes, the same wardrobe and sets, and continuity across cuts, which I wrote about in 2023 as the grammar of film. Today’s tools still struggle with that across even a few shots. Whoever makes character and scene consistency reliable at series length will unlock this whole category, and probably longer forms of storytelling after it.