AI video scripts and storyboards: narrative, shot design and prompts

A hand picks up a coffee cup, places it in a commuter bag, then lifts the bag to leave. This script describes a simple everyday action, but production requires decisions: What should viewers see first? When should you cut to a close-up? How does the cup travel from the table to the pocket opening? On which frame should a sound occur? Using a 30-second vertical product video, this article works through five stages to turn that creative goal into shots that can be generated, edited and reviewed.

This article is intended for creators who need to produce product short videos, brand content, and social videos. The case uses a fictional off-white coffee cup without labeling, and all scripts, time allocations, and prompts are original teaching plans; Part 5 has been supplemented with demonstration videos and actual production records generated and edited according to this scheme. When used for real products, they should first be replaced with verified product information and reference materials.

1. Overall structure: From the creative brief to the final video, a five-layer production structure is established

The creative brief for this example is: “Use the routine of getting ready to leave to help commuters remember this cup and encourage them to check its colors and specifications.” This gives you a way to judge whether a shot is necessary: holding the cup and packing it serve the brief, while a sudden spectacle of flying coffee beans would need a separate justification.

1.1 What belongs in each of the five layers

A shot list organizes production in words; storyboard images confirm compositions; an animatic places those images on a timeline for their intended durations. You can link all three in one project document, but each supports different checks. Before generating final footage, build an animatic with still storyboards and a temporary voiceover to catch pacing problems such as “the narration finishes before the action happens.”

Terminology and Process Reference: Runway:AI shot list: how to build a camera-ready plan . The five-layer structure in this article is a working method for short-form production, not a fixed format required by a model.

1.2 A creative brief you can use

 Project: Before You Go | 30-second vertical product video
Audience: Commuters who carry coffee daily and care about the appearance of their cup and how they use it.
Communication goal: Help viewers remember the product and commuting context, and encourage them to view its colors and specifications.
Core idea: Make taking a cup with you part of the daily routine of getting ready.
Factual boundaries: Show the cup body, closed lid, grip and insertion into the specified bag.
Verify the fit with the real cup and bag; generated visuals do not establish leak resistance or insulation performance.
Mood: A calm, orderly morning, expressed through movement speed, set design and sound.
Visual setting: Off-white cup, dark gray canvas bag, light wood tabletop, soft light from frame left.
Character: Only the same adult’s right hand and beige cuff appear.
Delivery: 9:16, 30-second finished video; choose output dimensions and frame rate for the publishing channel.
Sound: Voiceover, ambience and action sounds added in post-production; no on-screen dialogue.
Closing text: “Before you go, take this cup.” / “View colors and specifications.”
Available assets: Product photos from several angles, bag photos, hand and clothing references, and setting references.
To confirm: Actual cup and bag dimensions, approved brand copy, final voiceover and generation interface for each shot. 

“A premium feel” can begin a conversation, but the brief must translate it into concrete choices: show only the cup, bag and tabletop; limit the palette to off-white, dark gray and wood tones; keep hand movements gentle; and make key contact sounds clear. Photography, generation and editing can then follow the same decisions.

1.3 Keep global settings separate from shot-specific variables

Fixed items for the entire film Including the main subject's appearance, clothing, scene layout, lighting direction, and color relationships; Variables for each shot These include shot size, camera position, stage of the action and the duration used in the edit. Change only what the story requires in each shot. For example, cutting from a medium shot to a close-up of the lid changes the viewing distance, but the lid structure and right-hand grip should match the preceding shot.

Also need to be separated are "instructions for the human or host agent" and "prompts submitted to the generative model." The former can include audience, budget, shot usage, and acceptance requirements; The latter should highlight visible content and movement. The production sheet retains complete decision-making and selects information related to the shot during generation.

2. Scriptwriting: break down the theme into narrative beats and audiovisual content

The story summary in this example can be written as: "Early in the morning, a commuter holds a coffee cup at the table, puts it into the side pocket of the bag, then carries the bag away; The image ends with a separate product display." It provides the starting point, action, result, and conclusion. Next, the order in which information appears needs to be arranged.

2.1 Define what changes in each beat

Narrative beats are the turning points when the audience's cognition, character situation, or emotions change. In this example, the sequence follows "recognize the scene — pay attention to the product — start using it — complete packing — prepare to set off — get the next step." Six beats can correspond to six shots or be combined based on the material; first determine the narrative responsibilities, then decide the number of shots.

Different themes require different advancement relationships. Product demonstrations can use "usage problem—operation process—visible results—action entry point"; tutorials can use "objective result—necessary steps—common mistakes—result review"; narrative short films can revolve around "character goals—obstacles—choices—consequences." These are all methods for drafting, and the final order should be determined by specific expression.

2.2 Align visuals, sound, and screen text with audiovisual scripts

Short advertisements and commentary videos are suitable for audiovisual scripts: visuals and audio are recorded side by side, supplemented by time and text layers. This example uses the table below. Here, the time is the final production plan; The length of the generated clip and the actual usage points should be determined separately after the material is completed.

Audiovisual script format reference: StudioBinder:Ultimate AV Script Template to Write Better Ads . The timings, copy and shot designs above are original examples created for this article.

2.3 Duration is determined by both action and actual voice acting

First, record a temporary narration once, listen for natural pauses in the sentence, then adjust the screen length. Do not evenly distribute the length of each sentence based on the number of Chinese characters, nor cram the entire narration into all shots. S03 and S04 deliberately leave room for the sound, aiming to help the audience clearly see the cause and effect of the actions.

In this example, 8 seconds are reserved for packing to facilitate demonstration of action breakdown; If the actual editing is clear in 5 seconds, you can compress this segment and redistribute other segments or adjust delivery time. If the required finished length is 30 seconds, check the total again after making adjustments. Don't just slow down the action mechanically to fill time, or play the same clip over and over.

After the script is completed, do you check the cuts: Does each segment bring new information? Is the narration merely retelling obvious scenes? Is the ending just a clear next step? If a shot is removed and the audience can still understand the story without loss, then that shot is worth reconsidering.

3. Storyboard design: Use the language of the camera to clearly capture movement, space, and continuity

“Put the cup in the bag” is not yet a production-ready instruction. The storyboard must also establish whether the pocket opening is visible, which side the hand enters from, whether the cup’s base is obscured, and what state the final frame leaves for the following shot.

3.1 What information must be provided for a storyboard card?

 Shot: S04 | Setting: Morning at the table | Narrative purpose: Clearly show the cup settling into place
Finished-video range: 12–20 seconds; planned usable length: 8 seconds
Inputs: Approved S04 first frame; product, bag and hand references
Shot size and camera position: Close-up from the front side, angled slightly downward; pocket opening and gripping hand fully visible
Starting state: The right hand grips the cup, its base directly above the pocket opening; the bag rests on the table
Subject action: Cup descends vertically → pocket opening widens slightly under pressure → cup settles securely → fingers release → right hand leaves the frame
Camera: Fixed; keep the hand, the cup’s entry point and pocket opening in focus
End state: The lid protrudes above the side pocket, the bag remains on the table, and the right hand is no longer touching the cup
Sound events: Fabric friction begins as the cup touches the fabric; a light contact sound marks settling
Edit continuity: Match the cup-base position at the end of S03 and the completed insertion at the start of S05
Review priorities: Settle before releasing; the pocket opening does not merge into the cup; cup count and shape remain stable 

This card preserves the information the production team needs to check. For the video model, condense it into a few action sentences and keep the review criteria in the production record. Adobe’s shot list guide likewise emphasizes shot numbers, shot sizes, camera positions, movement and sound, while reminding readers to keep the list easy to use. Reference: Adobe shot list guide .

3.2 What shot size, camera position and camera movement each contribute

A dolly-in and a zoom are different movements: a dolly-in changes the camera's position, while a zoom changes the focal length. If the goal is simply to emphasize the product, describe the intended framing change first, then choose the appropriate term. Focal length and aperture numbers can express intent, but a generative model usually cannot guarantee an exact reproduction of real camera settings.

Most shots in this case use fixed camera positions to make continuous movements easier to judge; When emotional driving is needed, a clear purpose is added for camera movement. Combining rotation, propulsion, zoom, and complex hand movements in one experiment makes it harder to pinpoint the cause of failure.

3.3 Actions must have prerequisites, processes, and completion states

"Smoothly inserting the cup" still omits the key condition. The complete action should describe the contact relationship: who holds whom, where the bottom of the cup is aligned, when the pocket mouth starts to bear pressure, and when can you let go. For object interaction, "settle steadily first, then release" is easier to judge than adding "natural smoothness."

Actions can be broken down into three stages: "Preparation—Execution—Result." If a segment includes opening a door, turning around, drinking water, and changing scenes, it should be split into multiple shots first. For the difficult action in this example, two shots can be used to show "starting to enter the pocket" and "releasing after settling down," but the editing must help the audience understand what is omitted in the middle to avoid the illusion that the cup appears out of thin air.

3.4 Continuity Records "What Happened"

When joining adjacent shots, compare the preceding shot’s end state with the following shot’s start: is the grip in the same place, is the cup base at a similar height, and does the movement keep the same direction? If changing camera position during an action, also match its progress so a hand that has already lifted does not lift again in the next shot.

When involving characters in conversation, walking, or eye contact, axes can help maintain the left-right relationship: first determine the direction of movement between people or the main movement, then plan the camera's position. Crossing axes can be intentional, but transitional shots or clear spatial explanations should help the audience reorient. Reference: StudioBinder's explanation of the 180-degree rule .

Give each reference a defined role: product references constrain appearance, character references constrain identity, motion references supply movement information, and the first frame establishes the starting composition. Record each asset’s purpose and version. Whether an image can serve a particular role depends on the actual interface; calling it a “reference image” in the prompt is not enough.

4. Prompt writing: Convert storyboards into image descriptions suitable for generation mode

For the same S04 shot, the production chart can be written very thoroughly, but the video prompt only needs to emphasize: "Right hand places the cup vertically into the side pocket, after the cup is stable, release the hand, and the hand withdraws from the frame; Fix the camera, and finally stop at the stable state of the cup." "The premise is that the first frame correctly presents the cup, hand, bag, and bag opening."

4.1 First select the input mode, then decide what prompts need to be added

MiniMax’s official guide distinguishes text-to-video, image start/end-frame and reference-based input paths. rikoko prioritizes MiniMax interfaces that meet the task requirements. Before production, still verify the connected version, supported reference types, duration and output specifications. Reference: MiniMax Video Generation Guide .

4.2 Storyboard image prompts: describe a plausible single still

The storyboard is suitable for following this order: Aspect ratio and shot size → subject and reference roles → spatial relationships → current pose → lighting and materials → composition and negative space . Use this as a checklist; the final prompt can be a natural, connected description.

 Vertical composition, close-up, camera slightly downward from the front side of the table. 
According to product references, the off-white cup body and closed lid are retained, while the dark gray canvas side pocket structure is retained according to the bag reference. 
The bag is placed on the right side of the light wood table, with side pockets opening toward the camera. 
The right hand, wearing beige cuffs, holds the middle of the cup, with the bottom stopping directly above the pocket opening, not yet entering the side pocket. 
The hand, bottom of the cup, and bag opening are fully visible; the background is simple, with soft light coming from the left side of the image. 

This image captures only the moment just before insertion. Do not ask one first frame to show all four stages—picking up, inserting, releasing and leaving—at once. If you need a four-panel storyboard, identify it as a board for human review, then prepare a separate first-frame image for each image-to-video generation. Do not submit a whole multi-panel image unless the interface explicitly supports storyboard input.

4.3 Video prompts: Clearly describe the subject's movement, camera movement, and the ending state

Text-to-video requires creating images, while image-to-video focuses on variation. Google's Veo guide provides methods for organizing photography, subjects, actions, environments, and styles; Runway's Gen-4 guide specifically emphasizes that when images already provide appearance, text should prioritize describing movement. Both are suitable for their respective input scenarios and cannot be mechanically merged into a longer and better universal template.

References: Google Cloud: Veo 3.1 Prompt Guide ; Runway:Gen-4 Video Prompting Guide .

 Text-to-video draft structure:
[Shot size and camera position]. [Subject] is in [setting and specific position], starting from [initial state].
First [main action], then [necessary follow-up action], and finally [end state].
The camera [one clear movement or fixed setup], with [necessary lighting and style].

Image-to-video draft structure:
Starting from the first-frame pose, [subject] performs [main action]; [contact object or environment] responds with [necessary response].
Then [action result]. The camera [movement], ending on [a state that can be checked]. 

The last method allows the expected action to be observed and evaluated, but still needs to check the generated results. If the finger releases first, it records a "placement order error"; If the cup body sticks to the fabric, it records a "contact relationship error". Specific failures are more likely to translate into the next revision than "poor results."

4.4 Parameters, prohibited items, and sound descriptions should be placed in the correct positions

• Aspect ratio, resolution and generated clip duration: Set in the parameters supported by the tool used. The prompts "4K" and "8 seconds" cannot replace output parameters; The exact cutting point is still determined by the timeline.
• Reference materials: Submit via supported material input and indicate the purpose. Writing only a local file name in the text does not mean the model has received the file.
• Negative Prompt: Handle according to the specific model. Runway's Gen-4 guide recommends using positive expressions; For example, write "fixed camera" to avoid repeatedly writing "do not shake." If other interfaces provide negative fields, fill them in according to their documentation.
• Sound: Dialogue, ambient sounds, and action sounds are only considered native generation requirements when the selected mode supports audio generation. Otherwise, they are considered post-production sound scripts.
• Brand Wording and Subtitles: Keep exact copy in an editable text layer added in post-production whenever possible. If a label is part of the product surface, still check its shape, direction and occlusion frame by frame.

Requirements such as “stable, undistorted, physically convincing” can define review criteria, but repeating them does not guarantee the result. Address problems through the first frame, contact pose, motion complexity, camera occlusion and the usable portion of the clip.

5. Complete example: six shot prompts, sound editing and final review

Production Record of This Project: Following the feedback “the pacing is too slow; use English,” the player now shows a 15-second English cut. The six shots last 1.5, 2, 1.5, 4, 3 and 3 seconds: the wait after gripping the cup was removed, shot three no longer uses slow motion, insertion was compressed at 1.5× speed, and lifting the bag and leaving was edited at 1.25× speed. The closing frame retains 3 seconds for reading. The English voiceover, subtitles and closing copy were rewritten together. The original 30-second script remains below to compare the planned script with the final edit. Part of the person’s body still crosses the foreground in shot five, although no face appears; strict compliance with the “hands only” brief would require another revision. This fictional product example does not verify real cup-and-bag compatibility or product performance.

Next, the first four parts are organized into a set of materials that can be drafted. First, prepare and check four types of references: product appearance, bag structure, right hand and beige cuffs, desktop scene. The following "image prompts" are used to prepare the first frame of each take; "Video prompts" are used for the confirmed first frame's image-to-video format. Sound is uniformly produced according to the post-production plan of Part 2.

Before use, actually check the cup bag size relationship. This paper specifies that cups can fit into designated side pockets is a case premise and cannot be used as proof of true fitting. Each input should include the material required for that shot; "Reuse reference" is a production convention and does not mean the model will automatically remember the previous task.

5.1 First frame and video prompts for six shots

S01 | Establish the scene | Finished timeline: 00–03 seconds

 Image prompt:
Vertical medium shot. A light wood tabletop fills the lower half of the frame. An off-white cup stands upright on the left with its lid closed.
A dark gray canvas bag sits on the right, its side pocket facing the camera. Match the product and bag references respectively.
Simple background, soft light from frame left, and natural shadows on the table.

Video prompt:
The camera slowly moves toward the table while the cup and bag stay in place. Keep the movement subtle.
The cup and side pocket remain fully visible; the camera settles smoothly to a stop. 

Clip selection and review: Confirm the spatial relationship between the left cup and the right bag; both the cup lid and the bag opening can be recognized. If a slight push disrupts the product's shape, it can be changed to a fixed frame; The opening does not rely on complex camera movements.

S02 | Product and grip | Finished timeline: 03–07 seconds

 Image prompt:
Vertical close-up. An off-white cup stands on a light wood table with its lid closed, matching the product reference.
A right hand with a beige cuff is to the left of the cup; the fingers are close but have not yet gripped it.
The dark gray bag is visible out of focus in the background on the right. Soft light from the left; clear edges on the lid and hand.

Video prompt:
The right hand approaches from frame left, and the fingers gently close around the middle of the cup.
The cup’s base stays in contact with the table. Fixed camera; end on a stable grip. 

Clip selection and review: Check contact points and finger anatomy. At the end of S02, the cup has not yet lifted. Place the voiceover “Take your familiar cup” in this segment and leave a short, natural pause after the action.

S03 | Move above the pocket | Finished timeline: 07–12 seconds

 Image prompt:
Vertical medium close-up. The right hand already grips the off-white cup on the left of the table; the base still touches the tabletop.
The dark gray bag sits on the right, and the entire path between the cup and side-pocket opening is visible.
Use the approved product, hand, cuff and setting references, with soft light from the left.

Video prompt:
The right hand lifts the cup from the table and moves it steadily toward frame right.
The cup stays upright and stops directly above the side pocket, with a small gap between its base and the opening.
The camera remains fixed, and the hand keeps its grip on the cup. 

Clip selection and review: Pay special attention to the position relationship between the cup and the bag; the cup must not pass through the strap. The final pose of S03 can serve as a basis for the S04 first frame; When switching to a close-up, reconfirm the grip position and the height of the cup bottom.

S04 | Insert and release | Finished timeline: 12–20 seconds

 Image prompt:
Vertical close-up, viewed from the front side at a slight downward angle toward the dark gray bag on the right of the table.
A right hand with a beige cuff grips the off-white cup, whose base is directly above the side pocket.
The opening faces the camera. The hand, cup base and pocket are fully visible; appearance and grip match the preceding shot.
Soft light from the left; keep the cup and pocket opening in focus.

Video prompt:
The right hand slowly lowers the cup vertically into the side pocket; the opening widens slightly as the cup enters.
Once the cup settles securely, the fingers release and the right hand exits the frame. Keep the camera fixed.
End with the cup stable and its lid protruding above the pocket opening. 

Clip selection and review: Check each stage: insertion, settling and release. If the interface’s supported clip length cannot accommodate the complete action plus editing handles, revise the storyboard or choose a compatible mode. Simply writing a duration in the prompt will not change the supported length.

S05 | Lift the bag and leave | Finished timeline: 20–25 seconds

 Image prompt:
Vertical medium shot. The dark gray bag sits on the right of the light wood table; the off-white cup is securely in the side pocket with its lid visible.
A right hand with a beige cuff grips the bag’s handle; the bag’s base still touches the table.
Leave room for an exit toward frame right, with lighting consistent with the preceding shot.

Video prompt:
The right hand lifts the handle. The bag slowly rises from the table, then moves toward frame right and exits.
The cup moves with the bag, remaining in the side pocket. Keep the camera fixed and end on the empty tabletop. 

Clip selection and review: The cup must move with the bag. You can omit irrelevant movement between the release in S04 and the grip on the strap in S05. Connect the action with the sound of the strap taking the weight, as the voiceover says: “Ready? Let’s go.”

S06 | Standalone product shot | Finished timeline: 25–30 seconds

 Image prompt:
Vertical product shot. An off-white cup stands upright on a flat surface against a warm light gray background.
Match the approved product reference closely, with the lid closed.
Place the cup on the right and leave a clean, unobstructed text area on the left.
Soft light from the left shapes the cup’s outline; contact shadows look natural.

Video prompt (if subtle movement is needed):
The camera moves very slightly toward the product. The cup stays upright and still, and the text area on the left remains unobstructed.
Come to a smooth stop with the product outline clearly visible. 

Clip selection and review: This is a standalone ending screen for the ad. If product details have already been confirmed by static images, you can also slightly scale the static images in the editing to reduce visual changes caused by regeneration. Formal text is overlaid in post-production, checking for small screen reading and platform interface obstruction.

5.2 Assemble six segments into one video

1. Start by arranging the timeline with storyboard images and temporary narration. Arrange the shots at 3, 4, 5, 8, 5 and 5 seconds, for a total of 30 seconds. Check the cause-and-effect sequence of picking up the cup, inserting it and lifting the bag before producing final assets.
2. Record both generated duration and the duration used in the edit. In this example, each number represents the required footage. Select the clip length supported by the interface and reserve trim allowance before and after key actions; The actual amount generated is the basis for budgeting.
3. Test a key action such as S04 first. Once the still first frame passes review, prioritize testing the interaction between cup and pocket. If the action needs to be split into separate shots, revise the storyboard and timeline early, then settle the transitions to adjacent shots.
4. Match actions and directions during editing. Match the grip from S02 to S03, the position above the pocket opening from S03 to S04, and the completed insertion from S04 to S05. Adjust in and out points if movement repeats. For contact or identity errors, return to the asset-production stage.
5. Sound is positioned according to the event location, and text is positioned according to real speech. Cloth sounds synchronized with contact; Subtitles are calibrated based on the final dubbing. When voiceover appears, music is appropriately lowered, and action sounds and speech are not mutually masked.
6. Check the reading time at the end. S06 main sentences and action prompts should use clear hierarchy and contrast. Subtitles, safe zones, and volume requirements should be checked according to the actual release channel; templates from other platforms cannot be directly copied.

With rikoko MCP or another toolchain, use this material as production instructions for the host agent: organize the brief, storyboard and reference roles; preview each shot’s prompt and check generation requirements; then review the actual footage. The host must judge the visuals: a tool returning “task successful” only confirms execution. This is a recommended workflow; the specific tools and modes depend on what is currently connected.

5.3 Revise the failing stage instead of repeatedly remaking the whole video

Retain specific records of each modification, for example: "S04-v2: First frame unchanged, only the action changed to 'Landing steadily and then release'; Contact order improved, but the cup lid still deformed. "Prioritize fixing one problem at a time, check if it solves it, then proceed to the next problem. Shots that have already passed retain the original material and usage range for easier reassembly.

5.4 Before delivery, check the finished film using these six items

• Expression: Watching for the first time allows you to understand what the characters do and remember the main product information.
• Fact: Text, narration, and presentation do not exceed the verified product capabilities and adaptability.
• Identity: Products, hands, clothing, bags, and lighting are consistent in continuous shots.
• Action: The key contact action is complete and convincing, with no repeated movement, sudden direction changes or object jumps between shots.
• Picture and sound: Clear narration, synchronized action and sound, accurate subtitles, and ample reading time at the end.
• Delivery: Duration, aspect ratio, dimensions and file format meet the requirements; the project, asset versions and revision records are traceable.

To reuse this method, replace the brief in Part 1, then rewrite the information sequence in Part 2. Redesign shots and prompts for the new content. The reusable element is the decision process from objective to image to review, not just a paragraph from the cup example.