Script to storyboard: turn a script into a shot list
A practical method for going from a written script to a timed shot list with camera notes — plus how to automate the tedious half of it.
The reliable order is cut the script into information points first, assign seconds to each point, then assign shot size and camera note. Tools that jump straight to "generate images" skip step one, which is why their boards look good and cut badly.
Why "one click, one storyboard" disappoints
Ask an AI model to turn a 600-word script into a storyboard and you usually get one of two failures: a wall of prose describing shots, or eight beautiful frames that each last twelve seconds. Neither is a storyboard. A storyboard is a timing document — its job is to prove the episode fits its runtime and that information arrives at the right pace.
The missing step is segmentation. A script is written in sentences; a video is built from information points. One sentence can carry two points ("she opens the letter and starts crying"), and one point can span three sentences. Until the script is cut at information boundaries, no shot count or duration is meaningful.
The five-step method
- Cut by information pointOne point per shot. If a sentence carries two changes (action + reaction), split it. A 30-second script usually yields 8–12 points; 60 seconds yields 14–20.
- Assign seconds before anything else2–4 seconds per point is the working range. Under 2s the viewer cannot read the frame; over 4s the pacing sags. Dialogue-heavy points can stretch to 5s.
- Check the total against the target runtimeIf the sum overshoots by more than 15%, cut points — do not shave a tenth of a second off twenty shots. Uneven trimming is what makes an edit feel nervous.
- Choose shot size per pointWide establishes, medium advances, close-up carries attitude, insert emphasises. Never run three consecutive shots at the same size; the sequence reads as static.
- Write the camera note lastOne instruction per shot: slow push in, handheld follow, static, tilt up. Two movements in one shot is a promise the shoot cannot keep.
How many shots for a given runtime
Shot density is set by genre, not by taste. Action and comedy cut fast; drama and documentary let shots breathe. These ranges come from measured average shot lengths in released work and hold up well for social video.
| Runtime | Slow / drama | Standard | Fast / action, ads |
|---|---|---|---|
| 15s | 4 shots (avg 3.8s) | 5 shots (avg 3.0s) | 7 shots (avg 2.1s) |
| 30s | 8 shots | 10–11 shots | 14 shots |
| 60s | 14–16 shots | 20 shots | 26–28 shots |
| 90s | 20–22 shots | 28–30 shots | 38–40 shots |
If your board lands outside these counts for the genre you are making, the script is either too dense or too thin for the slot — fix it there, not in the edit.
A worked example: 30 seconds, 10 shots
Below is a real shape for a 30-second vertical piece. Note that durations are not round numbers and that no two consecutive shots share a size.
| # | Time | Picture | Line / info point | Size | Camera note |
|---|---|---|---|---|---|
| 01 | 0–2.8s | Rain on a window, city blurred behind | (no dialogue, one sound cue) | Wide | Static, slight rack focus |
| 02 | 2.8–5.4s | A hand pins an envelope to the table | "Three years. I never opened it." | Insert | Top-down, static |
| 03 | 5.4–8.6s | Her face, half-lit | "Today it goes out." | Close-up | Slow push in |
| 04 | 8.6–12.0s | Walking to the door, coat half on | Interior monologue, 1 line | Medium | Handheld follow |
| 05 | 12.0–14.2s | Hand hovers over the mailbox slot | (beat, music drops) | Insert | Static, shallow depth |
| 06 | 14.2–17.8s | She steps back, looks at the street | "What if he wrote back?" | Medium wide | Slow tilt up |
| 07 | 17.8–20.4s | Envelope edge, thumb pressing the seal | (no dialogue) | Insert | Static |
| 08 | 20.4–24.0s | Envelope dropping into the slot | "Sent." | Close-up | Static, 120fps feel |
| 09 | 24.0–27.2s | Empty street, she is gone | (ambience only) | Wide | Static |
| 10 | 27.2–30.0s | Title card over the wet pavement | Episode hook for next part | Graphic | Static |
Ten shots, 30 seconds, no shot longer than 3.6s, and every information point has its own frame. That is what a board has to prove before anyone draws a single image.
Camera note vocabulary that survives a shoot
| Note | What it means | Use it when |
|---|---|---|
| Static | Locked frame, no movement | Information is in the dialogue, not the frame |
| Slow push in | Dolly or zoom toward subject, under 10% of frame | Attention should tighten on one face or object |
| Handheld follow | Operator moves with the subject | Urgency, documentary feel, walking beats |
| Tilt up / down | Vertical rotation on a fixed point | Reveal height, scale, or a second subject |
| Rack focus | Focus moves between two planes | Two subjects, one decision |
| Whip pan | Fast horizontal blur into the next shot | Comedy beats, hard transitions only |
| Empty frame | No subject for 1–2s | Time passing, or a deliberate breath before a turn |
Doing it by hand vs automating the cut
Manually, this takes 40–90 minutes per episode at 30 seconds, most of it spent re-typing dialogue into a table and re-adding the seconds after every revision. The mechanical parts — splitting sentences at information boundaries, allocating durations that sum to the runtime, numbering shots, formatting the table — are exactly what a script should do.
ClipMynt's storyboard studio takes a pasted script, splits it into scenes and shots at the pacing you pick (slow, standard, fast), assigns durations that add up exactly to your runtime, and outputs a shot list with time ranges, shot size, camera note and dialogue. You edit any line and the timing recalculates. Export as Markdown or CSV for the shoot, or take the shot list straight into an image or video model.
What to check before you generate any images
- Does the total equal the runtime? A board that sums to 34 seconds for a 30-second slot will be cut in the edit anyway.
- Is the first shot under 3 seconds? Vertical feeds judge you before the fourth second.
- Does every information point have exactly one shot? Two points in one frame means one of them is wasted.
- Are shot sizes alternating? Three mediums in a row is the most common amateur signature.
- Is the last shot a turn, not a summary? End on a question the next episode answers.
What the tool hands back: the Storyboard Five
The method above is what a boarder does by hand. Run the same text through the studio and one paste returns five linked documents on a single timeline:
- Shot list — one row per shot: number, time range, size, angle, camera move, duration, frame, action and line. Markdown or CSV, ready for the AD.
- Image prompts — one English prompt and one keyword block per shot, written for Midjourney, Kling, Jimeng or Seedream, with lens, light direction and time of day already inside them.
- Dialogue and subtitles — each line cut to the seconds it needs at about 2.5 words per second, on accumulating timecodes, with the whole film exportable as an .srt file.
- Sound design — BGM direction, sound cues and mix notes per scene, so the recordist is not guessing from the picture alone.
- Production sheet — location, props, wardrobe, lighting plan, camera package, cast count and estimated shooting hours per scene.
The five stay consistent because they are generated from the same shot plan: change one shot's duration and its dialogue budget, its image prompt and its production row all follow.
Adapting a novel chapter, not a screenplay
Most text that reaches a storyboard tool is prose, not a script. A screenplay already tells you where the cuts are; a novel chapter tells you what someone thought. Three conversions do most of the work:
- Interiority becomes action. "She did not believe him" is unfilmable. Pick the visible tell — the pause before she answers, the hand that stops halfway to the cup — and board that. The line can carry the meaning; the frame carries the evidence.
- Narration stays narration, or it breaks the scene. First-person prose has no speaker on set. Keep those beats as VO with the speaker marked VO, and let the pictures run underneath. Turning a narrator's aside into dialogue makes characters say things people never say out loud.
- One paragraph often holds three scenes. Prose compresses time in a sentence ("three days later, the warehouse smelled of rain"). Each time jump, location change or reversal is a scene boundary — cut there first, then board the scenes.
A useful test: read the board without the source text. If a reader can follow who wants what and what changed, the adaptation worked. If they need the novel to understand a shot, that shot is still prose.
Frequently asked questions
Can the same method handle a novel chapter instead of a screenplay?
Yes. Cut at time jumps, location changes and reversals, turn interiority into visible action, and keep first-person narration as VO rather than dialogue. A novel chapter usually needs more scene cuts than a screenplay for the same runtime.
Can AI turn a script into a storyboard automatically?
Yes, for the planning half. A script-driven tool can split the text into scenes and shots, assign durations that sum to your runtime, and output a shot list with time ranges and camera notes. The visual half — generating consistent frames — still needs a model and a locked character description.
How long should each shot be?
Two to four seconds for most social video. Under two seconds the frame cannot be read; over four the pacing sags. Dialogue-heavy shots can stretch to five, and a deliberate empty frame can run six if it is doing structural work.
How many shots does a 60-second video need?
About 14 to 16 for drama, 20 for standard pacing, and 26 to 28 for fast cuts or ads. The count follows the genre, not personal preference.
What columns does a shot list need?
Minimum five: shot number, time range, picture description, dialogue or information point, and camera note. Add shot size and location if more than one person shoots it.
Do I need to draw the frames?
No. A timed table is already a usable storyboard for planning. Frames help when you are handing off to a cinematographer or feeding an image model; they are not required to prove the structure works.
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.