Vertical Drama Cliffhanger Types: 6 Episode Endings, Where to Cut, and a 12-Episode Rotation

Who this is for, and what it helps you decide

For writers turning in one to three vertical drama episodes a day, novelists adapting web-fiction chapters into short drama, and teams building script-tagging or breakdown tools. By the end you should be able to label the ending of each of your episodes, see whether your first 12 episodes keep repeating one move, and fix endings that cut a beat too late.

Current as of November 9, 2026 (Asia/Shanghai); all seven public pages were read that day. For how episode 1 opens, see our episode 1 cold open guide; for the single ending before the paywall, see the paywall cliffhanger guide. This piece covers the endings of every episode in between.

How seven public guides classify endings (facts)

Read this with care. All seven publishers sell AI video, AI screenwriting or script tools, and each article also promotes its product. None gives data on which kind of ending retains viewers better. ScreenWeaver says platform producers and an InkTip writers' guide use the same four shapes; we did not read the InkTip guide and do not paraphrase it. The reveal is the only type all seven list (some split it into identity, evidence and betrayal). Threat or danger, choice, and arrival or intrusion each appear in five.

The six merged types: where the last shot stops and how the next episode opens (our synthesis)

The table merges the seven lists. The examples are our own inventions, not taken from any series.

Two notes. Versely and Character App list the unanswered question as its own type; we treat it as a cutting technique, because any of the six can stop on a line that gets no reply. VerticalClap's faked death goes under threat, its physical evidence and revealed betrayal under reveal, and ShortDrama Studio's emotional turn (he finally says her real name) under reversal, as a flip in the relationship. Character App's return-of-the-thing (the ledger, the ring, the phone comes back) goes under arrival, with an object entering instead of a person; its visual sting, an ominous image carrying no information, reads more like a horror cutting choice and is not listed separately.

Where to cut: the one point nobody disputes

• Cut as the action starts, not after the result. Versely puts it as cutting on the setup of a beat, never the payoff: cut as the door starts to open, not once it is open, even mid-sentence. microdra.ma suggests finding the moment the viewer understands the new problem in the edit; it is often one reaction earlier than your first instinct.
• An ending and the next opening are one beat. ScreenWeaver calls this the pair rule: episode 12 ends on the door opening, episode 13 opens on who is in it. Its most common structural mistake is ending on the reveal and the reaction, then opening the next episode on a new scene. The reaction is the next episode's hook. Hold it back.
• Answer a small question, then raise a more urgent one. microdra.ma's example: she gets the envelope back (small question answered), opens it and finds her own signature on a contract she has never read (new question). Withhold everything and viewers stop trusting the storyteller.
• End on a face. ShortDrama Studio and VerticalClap both recommend a close-up reaction as the final image, because on a phone a face reads faster than a wide shot.

Disagreements and unsupported claims

• Which type belongs on the paywall episode. ScreenWeaver argues the last free episode should end on a revelation rather than a deadline or an intrusion, since a revelation creates a specific question and specific questions convert. ShortDrama Studio only says to save the strongest ending for the paywall. Axis, the production company quoted in our paywall guide, argues the paywall should come before the first big reveal. None of the three shows data.
• How many turns fit in one episode. Invideo says a 60-second episode has room for exactly one reversal and two would cancel each other out. ScreenWeaver puts the single turn about 60% of the way in, and separates it from the ending: the turn is a change the character makes, the cliffhanger a change the world makes to them.
• Outside disasters. VerticalClap says earthquakes and accidents land weaker than conflict between people. That is experience, not evidence.
• Rotation. ScreenWeaver says that if every episode ends on a revelation, by episode 15 the reveals carry no weight; ShortDrama Studio also recommends rotating. "Episode 15" is an illustration, not a measured figure.

A rotation for the first 12 episodes (editorial method, untested)

Assuming the app locks from episode 11, which is common on the main apps (see the paywall guide):

Three rules: never use the same type in two consecutive episodes; before the paywall, use a reveal no more than once every three episodes and save the biggest for the paywall; every ending should fit in a one-sentence question, and if it does not, the episode has no ending. Add "ending type" as a column in your beat sheet and log it the same way when you break down other shows.

Checked against series we have broken down (our labels)

These labels rely on the official episode synopses and third-party recaps used in our earlier breakdowns, not on a frame-by-frame viewing: - Agent of Shadows: episode 3 jumps eight years and an enemy lets slip a surprise, a reveal whose answer (he has a daughter) waits for episode 4; in episode 8 the lead is about to confess when a family turns up at the door, an arrival plus an unanswered question; episode 10 ends with a rival's sudden change of tone on the phone and the heroine frozen, a reversal sitting just before the episode 11 paywall. - Deny Me, Dragon King: the free run ends with the child calling out "Daddy" in front of the whole ball, a reveal and an arrival stacked together, heavier than either alone and in line with saving the strongest ending for the paywall. - The Betrayed Wolfless Is The Lycan King: the free run ends with the lead refusing compensation and vowing to take back everything. That is a vow, which fits none of the six well; choice is closest. The types are a tool, not a rule, and the goal-setting vow common in revenge stories deserves its own note.

Five questions for every episode before you submit

• Which type is this ending, and is it the same as the last episode's?
• Does the last shot stop as the action starts, or after the result?
• Is the character's reaction saved for the next episode's opening?
• Which small question did this episode answer, and is the question at the end more urgent?
• Is the threat, the newcomer or the evidence something the viewer already cares about?

What this means for products and topic choices (editorial inference)

• For tool builders: ending type is a structured field you can tag. A script-doctor or breakdown tool could label each ending with one of the six and flag "three reveals in a row" or "episode 10 is not the strongest ending", which is more specific and easier to check than a generic suspense score.
• For picking premises: stories with built-in public stages produce good endings more easily. Weddings, funerals, will readings, board meetings and mate-selection balls all let you end on someone walking in or a sheet of paper. When pitching, count how many such scenes the premise offers.
• For web-novel adaptation: a chapter-end hook and an episode ending do the same job, but a chapter can stop inside a character's head and an episode has to stop on something you can film. When adapting, ask of every chapter what the last shot is; if you cannot film it, rewrite the ending.
• For AI production: Invideo and Character App both treat the ending as its own shot instruction ("close on her hand"). When generating with AI, the final shot is worth producing separately, in several versions, before choosing.

What remains unclear

• All seven sources are tool-company blogs and none publishes retention or payment data; there is no public evidence on which type of ending works best.
• We did not read the InkTip guide that ScreenWeaver mentions.
• The merged six types, the rotation and the series labels are our editorial method, untested, and not any platform's submission standard.
• Series examples rely on official synopses and third-party recaps; the actual shots may differ.
• Episode length limits how many turns you can fit; see the episode duration guide. For terms, see the short drama glossary.

Sources

Accessed November 9, 2026 (Asia/Shanghai): - ScreenWeaver (2026-09-03), "Short Drama Episode Structure: Hook, Turn, Cliffhanger": https://www.screenweaver.ai/blog/short-drama-episode-structure - microdra.ma (2026-10-26), "How to Write a Cliffhanger for Vertical Drama Episodes": https://www.microdra.ma/blog/how-to-write-a-cliffhanger-for-vertical-drama - Versely (2026-08-19), "Cliffhanger Engineering for Serialised Verticals": https://www.versely.studio/blog/cliffhanger-engineering-for-serialised-verticals - Invideo (2026-07-15), "Writing Hooks and Cliffhangers for AI Micro-Drama": https://invideo.io/blog/ai-micro-drama-hooks-cliffhangers/ - Character App (2026-09-13), "How to Write a Vertical Drama Cliffhanger": https://www.character.app/guides/how-to-write-a-vertical-drama-cliffhanger - ShortDrama Studio (2026-10-15), "50 Short Drama Hooks and Cliffhangers (With Examples)": https://shortdrama.studio/short-drama-hooks-and-cliffhangers - VerticalClap (2025-05-10), "How to write an irresistible cliffhanger for your micro-drama": https://verticalclap.com/en/blog/comment-ecrire-un-cliffhanger-micro-drama

Related research & methods

How to Write the Episode 1 Cold Open for a Vertical Drama: The First 3 Seconds

How to Write a Vertical Drama Paywall Cliffhanger: Where Apps Lock Episodes and How to End the Last Free One (2026)

How to Break Down a Short Drama: A 12-Column Beat Sheet for Hooks and Paywall Episodes

Agent of Shadows, Episodes 1–10 Breakdown: How WWE's First Microdrama Builds to Its Paywall

Deny Me, Dragon King: Opening Breakdown of the Free Episodes (1–12)