Short drama storyboard: a 90-second vertical episode, shot by shot
Micro-dramas live or die on the first three seconds and the last two. Here is the shot density, the beat sheet and the framing rules that keep a vertical episode watchable to the end.
A 90-second vertical episode wants 26 to 30 shots at 3.0–3.5 seconds average, a hook resolved inside the first 5 seconds, a turn every 10–15 seconds, and a cliff on the final frame. Anything slower is a short film; anything faster becomes noise.
Why micro-drama pacing is not short-film pacing
A short film earns attention over minutes; a vertical episode is swiped away in one. Three structural consequences:
- The first shot carries the premise, not the setting. Open on the conflict already in motion — a hand, a door, a sentence mid-argument. Establishing shots belong at second 6, not second 0.
- Information arrives in beats, not scenes. Every 10–15 seconds a new fact lands. If two beats pass without new information, the viewer leaves.
- The last frame is a question. End on the reveal, not the resolution. The next episode exists because this one withheld something.
Shot density for vertical drama
| Runtime | Shots | Average shot | Beats | Shape |
|---|---|---|---|---|
| 60s | 18–22 | 2.7–3.3s | 4 | Hook, setup twist, reversal, cliff |
| 90s | 26–30 | 3.0–3.5s | 6 | Hook, setup, twist, escalation, reversal, cliff |
| 120s | 32–38 | 3.2–3.8s | 7–8 | Adds one escalation and one false resolution |
Series shot at 4+ seconds average almost always read as slow in the feed, no matter how strong the acting is.
A 90-second beat sheet
| Beat | Seconds | Shots | What must happen | Failure to avoid |
|---|---|---|---|---|
| 1 Hook | 0–5 | 2–3 | Conflict already running; one line of dialogue that raises a question | Opening on scenery or a logo |
| 2 Setup | 5–20 | 5–6 | Who wants what, and what blocks it | Explaining backstory in dialogue |
| 3 Twist | 20–35 | 5–6 | New fact changes the meaning of beat 2 | Twist that does not cost the hero anything |
| 4 Escalation | 35–55 | 6–7 | Pressure increases; hero acts and fails | Repeating the same obstacle |
| 5 Reversal | 55–75 | 5–6 | Hero changes tactic or learns the real stake | Resolving with luck or a third party |
| 6 Cliff | 75–90 | 3–4 | A new question opens; final frame holds the reveal | Explaining what happens next |
Worked board: 12 rows from a 90-second episode
Excerpt from a vertical drama board — the first beat and the cliff. Full boards at this density run 26–30 rows.
| # | Time | Picture | Line | Size | Camera |
|---|---|---|---|---|---|
| 01 | 0–2.2s | Close-up: a hand slaps divorce papers on a table | "Sign it. Today." | Close-up | Static, slight shake |
| 02 | 2.2–4.6s | Her face, unreadable, holding a pen | "You forgot one thing." | Medium close | Slow push in |
| 03 | 4.6–7.0s | Insert: the company seal on her desk | (no dialogue) | Insert | Static, rack focus |
| ... | 7–75s | Setup, twist, escalation, reversal — 18–22 rows | Beat-driven dialogue | Varied | Varied |
| 27 | 84.0–86.4s | Wide: the empty office, papers scattered | "Where is she?" | Wide | Handheld follow |
| 28 | 86.4–88.2s | Insert: a keycard with someone else's name | (beat) | Insert | Static |
| 29 | 88.2–90.0s | Close-up: her eyes, looking up off-frame | "You really want to know?" | Close-up | Static, hold |
Note the shape: two seconds at the open, two seconds at the close, and nothing in between that runs long enough to feel like a scene.
Framing rules for 9:16
- One face per frame. Vertical crops out context; two people in one shot means neither reads on a phone.
- Keep the eyeline in the upper third. Subtitle zones sit at the bottom, UI at the sides; the safe area is narrower than it looks.
- Cut on action, not on dialogue. Vertical audio is often played at low volume in public; movement carries the story when words do not.
- Inserts are your cheapest suspense. A keycard, a seal, a shaking hand — one-second inserts replace a paragraph of exposition.
- Never end on a wide. The last frame should be a face or an object the viewer wants explained.
Consistency across a 10-episode batch
Batch production fails on appearance, not on plot. Before generating anything, lock four things and reuse them verbatim in every prompt and every episode: hair and age, wardrobe per character, the two main locations, and the visual style sentence. Keep shot numbering continuous across the season so the editor can match boards to footage without guesswork.
ClipMynt's storyboard studio handles the planning half of this: paste the episode script, choose fast or standard pacing for the genre, and get a 26–30 shot board with time ranges that sum exactly to 90 seconds, camera notes per row, and Markdown or CSV export for the shoot. It also translates the full pack into English with identical shot numbers, which is what makes a bilingual team workable.
After the cut, run the finished episode through a timed breakdown and compare its real pacing with the board. The gap between planned and delivered seconds is the most useful number in the whole production.
The Storyboard Five, sized for a 90-second episode
A vertical episode is short but it is not small. The board that actually gets shot is five sheets — the Storyboard Five — generated from one plan:
- Shot list — 40–60 rows for 90 seconds at fast-cut density, each with time range, size, angle, move and duration.
- Image prompts — per shot, English, including the 9:16 framing rule and the character's locked description, so the same face survives every regeneration.
- Dialogue and subtitles — lines cut to 2.5 words per second, on accumulating timecodes, exportable as .srt for the editor's timeline.
- Sound design — per scene: BGM intent, cues, mix notes. Vertical drama leans on sound to carry cuts that have no coverage.
- Production sheet — location, props, wardrobe, lighting, cast and hours — the sheet that tells you whether ten episodes fit the shoot days.
Assets to lock before episode two
In a series, the board is cheap to regenerate and the look is expensive to re-establish. Three things should be locked once and reused across the batch:
- Character anchor sheets. One front-facing, evenly lit image per character, with the description written in fixed words (age, build, hair, wardrobe defaults). Every shot prompt references those words. Without it, the lead's face drifts by episode three and the audience drops out without knowing why.
- Empty location plates. A plate per set with no people in frame — the office, the stairwell, the rooftop. Shots are composed against the plate so the set does not reshape itself between episodes. Negative-prompt people out of the plate, or the model will seed a stranger in the background.
- A fixed title and cast block. The same names spelled the same way in every file. Consistency downstream (subtitles, exports, thumbnails) is a naming problem long before it is an artistic one.
Lock those three and a ten-episode batch becomes ten passes of the same pipeline. Skip them and every episode is a pilot.
Frequently asked questions
How do I keep the same faces across ten episodes?
Lock one character anchor sheet per lead (front-facing, evenly lit, fixed description words) plus an empty plate per location, and reference both in every shot prompt. Without anchors the face drifts within two or three episodes.
How many shots should a 90-second short drama have?
26 to 30 at 3.0 to 3.5 seconds average. Below 24 the episode reads as slow in a vertical feed; above 34 the audience cannot track who is doing what.
How long should the hook be?
Two shots, under five seconds total, with the conflict already in motion. The premise has to be visible before the viewer's thumb moves.
Do short dramas need establishing shots?
One at most, and not first. Place it around second 6 to 10, after the conflict has been introduced.
Can AI storyboard a whole season?
Yes for planning: script in, timed shot list out, consistent numbering across episodes. Visual consistency still depends on a locked character description reused in every prompt.
What is the ideal shot length for vertical video?
Three seconds is the working average. Under two seconds only for inserts and reactions; over four only for a deliberate hold on a face at a turn.
Notice: guides are for studying public content and are not professional advice. Copyright stays with the original creators; commercial use requires permission. Results vary by category, script and platform — trust your own numbers.