Vertical short drama, one episode at a time
Short-drama apps took about $750M in in-app purchases in the first quarter of 2026, and the most profitable of them already ships AI series under its own brand. We measured six episodes shot by shot. The format is strict, which makes it easy to copy.
The format
Episodes run 1 to 2 minutes, vertical, with hard cuts only: no fades, no wipes. A shot lasts about 2 seconds, so 25 to 35 shots a minute, and about 60% are close-ups. A scene is one place and one beat for 10 to 20 seconds, then the story moves. Captions carry every word, so the episode works with the sound off.
The story loop
Every episode runs the same loop. A hook in the first 3 seconds: one line of status or threat. Then a humiliation, a shove, a theft or a public rejection every 10 to 15 seconds, each followed by a crowd reaction. The end is a silent close-up of one face for 3 to 5 seconds, then black. The episode ends on the face, never on the line. Across a season the episodes climb a ladder: a hidden identity, repeated humiliation, small wins, the identity revealed near the middle, then revenge and romance.
Cast and wardrobe
Three to six named characters: the heroine, the male lead, a rival, an older villain, an assistant. Wardrobe tells the viewer who is who: the rival wears red, the wronged heiress has a torn dress. In our own trailer, the café girl has a black apron and a cheap tortoiseshell hair clip, the rich student a charcoal overcoat and a steel watch, and the rival a red wool coat. Give each character one look and keep it.
Make episode 1 in ViewPress
Make the cast first. For each character, ask for a portrait and save it as a character: “A portrait of Yuna, 22, who works the counter at a café by a university gate: black apron over a grey t-shirt, hair tied back with a cheap tortoiseshell clip, a guarded look, soft morning light.” Then write the episode in the chat: “Episode 1, 75 seconds, vertical. A rich student mocks Yuna at the counter. At closing he comes back and offers her money to pretend to be his girlfriend. Hook: his first line. End on her face as she turns his card over.” Read the scenes on the strip before you make any frames. Then make the scenes one at a time, with each character named with @.
Dialogue and voices
Keep spoken lines short. A realistic talking scene holds its lip-sync best at about 7 seconds or less, so split a long line across two shots. Speech runs 2.5 to 4 words a second; size each scene to its words. Reactions, crowds and close-ups of objects (a phone, a ring, a business card) can be stills with a slow push-in. Give each character a voice on the project’s Voices page and use it in every episode. Voices that sound synthetic are the top complaint about AI drama apps.
Titles and posting
The title is the promise in five to eight words: “Her Empire of Billions”, “They Broke Me, Now They Beg”. The thumbnail is a movie poster: one couple or one woman, the title, an episode badge. Post one episode a day. The daily drop is what brings viewers back.
Frequently asked
How many episodes should a season have?
AI series on the apps run about 20 episodes. Start with 10 to 20, and put the biggest reversal at episode 6 to 10, where the apps put their paywall.
Can I narrate instead of using dialogue?
Yes. One of the top AI drama apps tells most of its first episode as narration over stills with a slow zoom, and lip-syncs only the confrontation. Narration over stills reads as a choice. Bad lip-sync reads as broken.