A useful animation storyboard specifies what changes in each shot and what must already be true when the next shot begins. Write those states before adding cinematic adjectives. Then create separate instructions for the starting image and for the movement that follows.
Dreamina can help turn approved artwork and prompts into short animated footage. The storyboard remains your plan for choosing inputs, reviewing candidates and deciding where to cut. A prompt requesting a precise action is an instruction to evaluate, not proof that the generated clip will perform it correctly.
Here is a complete proposed example: a small robot stops a wooden ball from rolling off a workbench. It uses four shots with an 18-second editorial target. The story, cards, prompts and diagram below are instructional examples; they have not been generated or tested in software.
- Find the visible change inside the idea
- Lock the geography before dividing the action
- Allocate time by information, not by sentence length
- Build a card for each shot, including its handoff
- Turn the cards into an animatic before polishing the images
- Move the approved plan into Dreamina
- Repair an overloaded shot by changing the board
- Deliver a board that another operator can use
Find the visible change inside the idea
The loose idea is “a helpful robot saves the day.” That describes a personality, but it does not tell you what to show. Make the problem visible: a yellow ball is rolling toward the right edge of a workbench. Make the intervention visible: the robot lowers a red safety bar across its path. Make the result visible: the ball meets the bar and stops before the edge.
The story needs no dialogue. The audience should understand it from the ball's direction, the edge, the bar and the robot's reaction. Music can add warmth, but it cannot supply missing cause and effect.
Before drawing, write the completion condition: the same ball rests against the lowered bar, with visible tabletop between the barrier and the drop. This excludes several attractive but wrong endings: the ball disappears, the robot catches a different ball, the bar becomes a wall, or the camera hides the edge before the outcome can be checked.
Choose the production route around that requirement. If broad action with reviewed results is sufficient, generate shot candidates in Dreamina. If the ball must follow an exact trajectory or contact must occur on a specific frame, use an editable animation setup for that shot. The animation workflow decision guide explains the difference between delivering footage and delivering reusable controls.
Lock the geography before dividing the action
Use a simple workbench, one robot, one ball and one bar. The camera stays on the same side of the action axis throughout the sequence. In every view, the ball approaches from screen left and the dangerous edge remains to screen right.
The robot stands behind the ball's path, farther from the camera. It holds the red bar above that path, ready to lower it. The bar will cross the path, rather than run alongside it. At its final position, its lower edge meets the tabletop and leaves clear table space before the drop.
Keep the design card short:
This plan avoids asymmetric clothing or decorative markings that do not serve the story. It does not avoid the important challenge: ball-to-bar contact. That is the moment to prioritize in review and the first place to consider a more controlled production route if repeated candidates fail.
Allocate time by information, not by sentence length
The four beats are discovery, intervention, contact and relief. Their durations are deliberately different:
Original planning schematic, not generated footage. The panels summarize shot states and timing; their simplified shapes are not approved character artwork or first-frame inputs.
The total is 4 + 5 + 4 + 5 = 18 seconds, with hard cuts and no added title card. Adding a two-second title would make the plan 20 seconds unless another shot is shortened. A transition that overlaps clips can also change total duration. Keep the editorial arithmetic separate from whatever durations the generation interface offers.
The first shot establishes danger without letting the ball reach the bar location. The second shot crops the moving ball out, so the audience can focus on the robot's intervention. The third supplies the missing physical event. The fourth gives enough time to recognize the outcome. This is story timing, not a physically calibrated simulation of the ball's speed.
If the available generation duration is longer than a planned shot, request enough usable movement and trim later. Do not ask a four-second minimum interface for a one-second generation just because the final insert will be short. A retained interval and a submitted clip duration are different quantities.
Build a card for each shot, including its handoff
Each card below gives the start state, action, end state, references and rejection conditions. The time ranges inside a prompt are local to that clip: “0–1 seconds” in the third shot means the beginning of that clip, not the beginning of the whole film.
Use the accepted robot image for identity, the accepted workbench layout for location and the shot's approved starting image for composition where the selected mode supports that role. Do not upload a labeled four-panel teaching diagram as though it were one scene's first frame.
Shot 1: Discover the danger — four seconds
Purpose and framing: A stationary wide view establishes the table edge, the left-to-right ball path and the robot behind it. Keep the edge visible from the start; revealing it only after the ball has stopped would weaken the reason for the intervention.
Start: Ball well to the left of the robot and bar location. Robot holds the bar raised. There is a visible gap under the bar. Both feet rest on the table; the robot looks forward.
Action and end: The ball rolls slowly right. The robot directs its eye toward it and becomes attentive. End before the ball reaches the barrier location. The bar remains raised.
Proposed starting-image prompt:
Create one wide storyboard image on a gray workbench with a clearly visible drop at screen right. A small copper robot with a square body, one dark oval eye, two arms and two feet stands behind the ball's path. It holds one red rectangular safety bar raised above the tabletop, across that path. One plain yellow wooden ball sits well to screen left of the bar. Show the full robot, the ball and clear tabletop between the bar location and the right edge. Soft illustrated surfaces, clear outlines, pale background, no text or extra objects.
Proposed motion prompt:
Use the approved wide image for composition and the accepted robot reference for identity. One four-second shot, stationary camera. From 0–2 seconds, the yellow ball rolls slowly toward screen right but remains clearly upstream of the raised bar. From 2–4 seconds, the robot notices the ball and looks toward it. Keep the bar raised throughout. End with the ball still left of the bar location. Preserve one ball, one robot, the tabletop and the right-hand edge. No cut or camera reversal.
Reject if: The ball reaches the edge, the bar lowers early, another ball appears, or the robot's reaction cannot be read. A gentle gaze shift is enough; a dramatic spin would complicate the next cut unnecessarily.
Shot 2: Lower the barrier — five seconds
Purpose and framing: A tighter view concentrates on the bar and the robot's arms. Keep the camera on the established side. The approaching ball is outside the left edge of this tighter frame, not removed from the story.
Start: The robot and raised bar match the end of the wide shot. The tabletop is visible below the bar, and the right-side safety margin remains understandable.
Action and end: The robot lowers the bar until its lower edge meets the table. End on a stable, complete barrier. Do not ask the robot to move the bar sideways and lower it at the same time.
Proposed starting-image prompt:
Use the accepted robot and workbench references. Create a medium-close view from the same camera side as the wide shot. Show the copper robot holding the same red rectangular bar raised over the gray tabletop. The bar crosses the established ball path. Crop the upstream ball out to screen left. Keep space visible between the bar and the table edge at screen right. Match the previous raised-bar pose and illustrated treatment. No added object, text or second character.
Proposed motion prompt:
One five-second shot with a stationary camera. From 0–1 seconds the robot holds the raised bar. From 1–4 seconds it lowers the same red bar straight down to the tabletop. From 4–5 seconds it holds the bar still with its lower edge meeting the surface. Preserve the robot's design, its grip and the workbench position. The ball remains offscreen to the left. No cut, sideways sweep, new barrier or camera move.
Reject if: The bar never meets the surface, passes through the table, changes shape or detaches from the hands. An appealing lowering movement with a ball-sized gap underneath cannot perform the story's job.
Shot 3: Show the contact — four seconds
Purpose and framing: A low insert shows the ball approaching the lowered bar and the safe strip of tabletop beyond it. The robot's face is unnecessary here; include enough of the grip or body to make the barrier relationship believable without filling the frame with competing detail.
Start: Bar already down. Ball just to its left with clear separation. Table edge stays farther right. The ball's size matches the wide view after allowing for the closer camera.
Action and end: The ball rolls right, touches the bar and stops. Finish with the contact visible. The bar does not move away and the ball does not pass through it.
Proposed starting-image prompt:
Create one low close-up on the same gray workbench from the established camera side. One yellow wooden ball is just left of the lowered red rectangular bar, with a small visible gap before contact. The bar's lower edge meets the tabletop and crosses the ball's path. The robot's grip holds the bar from behind. Beyond the bar, show clear tabletop before the drop at screen right. Preserve the accepted ball, bar and illustrated style. No extra ball, writing or camera-axis reversal.
Proposed motion prompt:
One four-second close-up, stationary camera. From 0–2 seconds the yellow ball rolls a short distance from screen left toward the lowered red bar. It contacts the bar and stops. From 2–4 seconds the ball remains resting against it. Keep the bar held firmly at the tabletop, the ball intact and the table edge unchanged. No bounce over the bar, disappearance, duplicate sphere, camera move or cut.
Reject if: Contact is hidden, the ball changes size abruptly, it passes through the bar, or it stops before reaching it. This is the causal proof of the rescue. Do not accept a vague shot because the preceding and following panels look correct.
Shot 4: Let the outcome register — five seconds
Purpose and framing: Return to the same wide geography as the opening. A calmer robot reaction gives the sequence an ending while keeping the saved ball visible.
Start: Ball resting against the lowered bar. Robot still holds it. The right-hand edge and intervening safety margin match the contact shot.
Action and end: The robot relaxes slightly and looks at the safely stopped ball. It keeps holding the bar. The final pose is stable; do not introduce a pickup, celebration jump or new departure.
Proposed starting-image prompt:
Recreate the accepted opening wide composition on the same workbench. The copper robot now holds the red bar lowered against the tabletop. The single yellow ball rests against its upstream left side. Keep the right-hand drop and visible safety margin beyond the barrier. Match the accepted robot, object sizes, lighting and illustrated treatment. The robot looks toward the stopped ball. No new prop or lettering.
Proposed motion prompt:
One five-second stationary wide shot. During 0–2 seconds, the robot makes a small relaxed body movement and looks at the yellow ball resting against the bar. During 2–5 seconds, hold the safe arrangement with subtle idle motion only. The robot keeps holding the lowered red bar. The ball does not roll again. Preserve the established workbench, screen direction and object count. No pickup, new gesture, camera move or cut.
Reject if: The ball starts moving again, the bar rises, the edge changes side or the robot's reaction alters the design. The story already has its action; the ending needs confirmation rather than another event.
Turn the cards into an animatic before polishing the images
Use rough rectangles and circles if necessary. Place the four panels in sequence for 4, 5, 4 and 5 seconds. Add temporary sound cues: a light rolling sound in discovery, a lowering movement in intervention, a small contact sound in the insert, then a calm ending. These are proposed finishing responsibilities, not audio already generated with the shots.
Watch the rough sequence silently first. Can someone identify the danger and the rescue? Then listen with temporary sound. Place the contact cue where the ball actually meets the bar, not merely at the start of the insert. Avoid using a loud impact to imply contact that the picture never shows.
Check adjacent cards in pairs. Discovery-to-lowering must preserve the raised bar. Lowering-to-contact must preserve the bar touching the table. Contact-to-settle must preserve the stopped ball. A close-up can hide part of the location, but it must not contradict the established screen direction.
If you change the wide view, update both the opening and ending cards. If the robot's grip changes, revisit the lowering and contact cards together. A storyboard is useful because it tells you which decisions depend on each other.
Move the approved plan into Dreamina
Prepare the robot master, a location layout and four individual starting images. Dreamina's Seedream 5.0 Pro guide documents AI Image, text or image upload, model and aspect/resolution settings, generation, review and download. Use the card's starting-image prompt for that stage. Inspect object count, bar height, ball position and framing before approving each image.
For motion, the documented Seedance 2.5 route uses AI Video and Omni references, with uploaded assets identified in the prompt. Assign each reference its intended job and use the matching motion instruction. Choose durations and settings available in your account. An Omni identity reference and a dedicated first-frame control are different roles; use the documented control for the mode you actually select.
The transferable principle is to supply clear, purposeful references. Runway's Gen-4 Image References documentation likewise describes references and input preparation, but its control names and limits belong to Runway. They do not establish Dreamina input limits or guaranteed continuity.
Save the exact approved inputs and prompts with each candidate. This makes it possible to distinguish a wrong starting state from an action that failed after a correct start. For a bounded real example of reference relationships across multiple shots, see the three-shot mascot case; it is a separate saved demonstration, not a result of this robot storyboard.
Repair an overloaded shot by changing the board
A tempting first prompt might say: “The robot notices the rolling ball, runs to the edge, catches it, smiles and waves while the camera circles.” That combines locomotion, interception, a hand interaction, facial reaction and camera motion. Even if a candidate is attractive, it is difficult to identify which event communicates the rescue or where it can be repaired.
The four-shot board removes the run, replaces the catch with a barrier and separates contact from reaction. If the ball still fails to stop convincingly, revise only the contact requirement's production plan:
That revision totals 16 seconds. The two-second insert is an editorial duration, not a claim about a generation minimum. If the brief still requires 18 seconds, spend the extra two on readable discovery or relief rather than inventing another action.
A cut from approach to a stopped state may make the story understandable, but it does not demonstrate a continuous physical collision. If showing that exact event is mandatory, retain the continuous contact requirement and produce it with direct animation controls. Do not label an omitted event as a successful generated action.
Deliver a board that another operator can use
Hand over the story sentence, design card, geography sketch, numbered shot cards, individual starting images and 18-second animatic. Mark which images are approved and which prompts remain proposals. Keep the local shot timings separate from the sequence timeline so the operator does not submit “9–13 seconds” as though the model were continuing a 13-second film.
After candidates exist, select usable intervals and check the cuts again. The animation assembly guide handles those editing decisions, including pacing, sound and final-file review. The storyboard's job is to make that edit possible: one clear change per shot, visible evidence of the result and a deliberate state passed across every boundary.
Begin with rough panels in Dreamina or a simple drawing tool. Approve the causal story and the handoffs first. Polished surfaces can improve the film later; they cannot explain why a ball that never met the barrier has suddenly stopped.