What a micro drama is, in four numbers
Micro dramas, also called short dramas, vertical dramas or Chinese short dramas, are serialized stories built for phones. An episode runs one to three minutes in 9:16 vertical format. It opens on a visual hook in the first three seconds, builds around one conflict, and ends on a cliffhanger that pulls the viewer to the next episode.
Google searches for "short drama" and "AI drama" roughly tripled worldwide between October 2025 and October 2026. Searches for the two largest dedicated apps, ReelShort and DramaBox, peaked over Christmas 2025 and have stayed high since. Four numbers show the scale of the format:
- 11 billion USD in global micro drama revenue in 2025, projected at about 14 billion USD in 2026, according to Omdia.
- 3.6 billion USD of that revenue outside China in 2026, with ReelShort holding 29 percent and DramaBox holding 21 percent of the non-China market, according to the MPA (reported by Screen Daily in August 2026). Both apps earn from coins per episode and subscriptions.
- 38 percent of China's top 100 micro dramas in January 2026 were AI titles, up from 7 percent a year earlier, according to The Next Web.
- 100+ AI micro dramas launched in a single batch on BULLET, the ZEE-backed micro drama app in India, in October 2026 (MediaNews4U).
The format works for the same reason soap operas worked: a cast you recognise, a problem that does not resolve in one sitting, and a reason to come back tomorrow. AI fits the format because vertical drama viewers care about the story hook and the cliffhanger more than big-budget sets.
The AI micro drama stack
To make a short drama with AI, production breaks down into three technical layers, a structure outlined by Justine Moore at a16z in July 2026:
- Foundation video models that generate the actual moving shots from image references, such as Seedance 2.5, Kling, Veo and Hailuo.
- Creation tools that turn a written premise into a storyboard, manage character references, and assemble the cut.
- Distribution platforms where viewers watch the episodes.
There is a catch in that third layer for solo creators. Apps like ReelShort and DramaBox commission and license studio series. They do not take open uploads from an individual with a laptop. For an independent creator, the real distribution layer is TikTok, YouTube Shorts and Instagram Reels, where the recommendation algorithm is the only gatekeeper.
On open social feeds, a vertical drama AI series lives or dies on one thing: whether viewers recognise the lead from one episode to the next. Studios solve this by hiring actors. A solo creator can solve it by starring as the lead. When your own face is the recurring character, your audience follows a person rather than an anonymous channel. In Doppelshot, the Video Studio is built around this exact setup: your selfie or AI twin plays the lead across a 60-second, ten-shot episode.
Step by step: genre, premise, ten shots
Making a 60-second episode that holds attention on TikTok, Reels or Shorts follows a strict structure.
1. Pick one genre and stay in it
Micro dramas are genre television. Pick one lane and make ten episodes before trying another. The genres that hold retention best on vertical feeds are revenge, CEO romance, secret heir, heist, werewolf, and deadpan comedy such as a house cat drama played like a serious soap opera.
2. Write one line of premise
Keep the episode premise to a single sentence with one conflict and one promise. For example: "My boss fires me on Monday, and by Friday I own the company." If the premise needs two sentences to explain, it belongs in two separate episodes.
3. Turn the premise into ten shots
A 60-second vertical episode is about ten shots of four to eight seconds each. A reliable beat sheet looks like this:
| Shot | Time | Job |
|---|---|---|
| 1 | 0 to 3 s | The hook: the most dramatic image of the episode, before any context |
| 2 to 3 | 3 to 15 s | Setup: who the lead is, where they are, what they want |
| 4 to 6 | 15 to 35 s | The turn: someone says or does the thing that changes the plan |
| 7 to 9 | 35 to 55 s | The reveal: the hero learns a secret or makes a decision |
| 10 | 55 to 60 s | The cliffhanger: an unanswered question that forces part two |
Write each shot as one physical action and one camera move with a clear end point, such as: "She slides the envelope across the desk, slow push in, ends on her hand." Put spoken lines only on medium shots and close-ups where the character's mouth is large enough to read clearly.
Continuity: the same face in every shot
Character consistency is where most homemade AI episodes fall apart. Across ten shots, the face drifts, the hair changes length, the jacket turns into a different coat, and the room rearranges itself behind the actors.
Working through weekly episodes in the Doppelshot Video Studio taught us five rules that stop the drift:
- Lock the lead and the outfit before rendering video. A single front portrait is not enough when a character turns around. Create a reference of your lead in the episode outfit from the front, three-quarter view and back. In Doppelshot, you can start from one selfie or build an AI twin from 10 to 20 selfies so your face stays steady.
- Lock each location with a set anchor. Rooms drift just like faces. Generate one master still per location with the lead standing in it, in the right outfit and lighting. Start every shot in that room from that anchor.
- Never chain a shot from the last frame of the previous video clip. Taking the final frame of shot two as the start frame of shot three looks like an easy shortcut for continuity. In practice, every video generation degrades slightly, and by shot six the face no longer looks like you. Start every shot from its own clean keyframe still generated from the set anchor.
- Repeat the person and outfit description in every prompt. Video models do not remember shot one when rendering shot four. State hair, eyes, build and exact clothing in every shot prompt, and check the first, middle and last frames of each rendered clip against the portrait.
- Check all ten stills before paying for video. Generating still keyframes is fast and cheap. Rendering ten video clips is the heavy step. Preview the ten still frames first, redraw any shot where the outfit or hand looks wrong, and only then run the video shoot.
For the video generation step itself, pick the engine that fits the story. Hailuo holds a reference face well and is a safe default. Seedance 2.5 keeps a reference person steady across multi-shot scenes and generates action sound with the picture. Veo 3.1 adds realistic lighting and ambient audio. In Doppelshot, each card runs on a default model, or you can switch engines in the Studio; the price shows before you run, and you can check how credits work on the pricing page.
Voices, music and captions
A silent vertical video gets scrolled past in two seconds. Once the ten video shots are ready, the audio and text layers turn separate clips into an episode.
In Doppelshot, dialogue, music and subtitles run on ElevenLabs. You assign a voice to each character from about 30 built-in voices, or clone your own voice so the lead sounds like you. Keep spoken lines short and place them on close-ups. Add a background score that matches the genre, burn in readable vertical captions for viewers watching on mute, and finish with a title card that says "To be continued" so viewers know a next part is coming.
When episode one ends, you do not have to rebuild the cast or rooms from scratch. In the Video Studio, the next episode opens directly on your previous cliffhanger in one tap, keeping the same cast, outfits and sets for episode two. Every finished episode carries a machine-readable AI marker in line with Article 50 of the EU AI Act, and you should also turn on the AI-generated content toggle when posting to TikTok, YouTube Shorts or Reels.
Ready genres you can start from
You do not need to invent a beat sheet from a blank page. Doppelshot includes ready episode cards where the ten-shot structure, pacing and camera moves are pre-built around your selfie or twin:
- Revenge and intrigue: Revenge, Secret heir, Heist, Wrongly accused, Family secret, Caught, and Desert deal.
- Romance and melodrama: CEO romance, where power, a contract and a held glance drive the ten shots.
- Supernatural and sci-fi: Werewolf, Ancient gods, Alien, and Baby dragon.
- Comedy and everyday stories: Cat drama (a soap opera about a house cat played completely straight), Stick figures, Home video, and Morning routine.
- Brand and creator spots: Product ad and UGC ad for 30-second commercial stories with your face and your own product.
- Kids cartoon packs: Forest friends, Space kittens, Dragon school, Bedtime stories, and Little climber, where a parent runs the account and the child appears only as a cartoon character.
FAQ
What is the difference between an AI micro drama and a short drama?
There is no difference in format. Micro drama, short drama, vertical drama and Chinese short drama all describe serialized vertical episodes of one to three minutes built around a three-second hook and a cliffhanger. An AI micro drama simply uses AI image, video and voice models instead of a film crew.
How much does it cost to make an AI micro drama episode?
Third-party estimates put an AI episode between 50 and 500 USD, and one published case study counts about 1,000 USD per episode for a three-person team, against 200,000 USD and more for a studio season with actors. On Doppelshot, you pay with credits per preview still and per video shoot with no mandatory subscription; the price shows before you run, and all packs and plans are listed under pricing.
Can I use my own face and voice in an AI short drama?
Yes, if you are an adult (18+) and use your own photo or have the explicit agreement of every adult shown. You can start from a single selfie or create an AI twin, and you can pick from about 30 voices or clone your own voice. Uploaded selfies and twins are stored in a private bucket, are never used to train Doppelshot's models, and can be deleted at any time from your account. Every paid result includes a commercial license with no watermark.
Can one person make a full vertical drama series?
Yes. By locking your character and set anchors once, you can write, preview the ten stills, render the video shots and add voices in one evening. When an episode finishes, the writer opens the next episode on your last cliffhanger while keeping the same cast, outfits and locations.
