Storyboarding an AI video before you render
How to storyboard an AI video: shot sizes, angles and moves, a panel template you can copy, and why drawing stills first saves credits and keeps shots matched.

The short answer
A storyboard is a sequence of still frames, one per shot, with notes on framing, movement, action and sound. For AI video, storyboarding means generating those stills first, approving them, and only then turning each one into motion. It is the cheapest place to catch mistakes.
Images cost less than video and render faster. If a frame is wrong as a still, it will be wrong as a clip. Fix the composition, the character and the light in the storyboard, and the video stage becomes animating decisions you have already made.
The language of a storyboard
Shot sizes, from far to near:
- Extreme wide shot: the setting dominates, the person is small. Establishes place.
- Wide shot: the whole body in its surroundings.
- Medium shot: waist up. Most conversation.
- Medium close-up: chest up. Emotion with some context.
- Close-up: the face. Feeling and reaction.
- Extreme close-up: eyes, hands, an object detail.
- Insert: a cutaway to an object that matters to the story.
Angles: eye level (neutral), low angle (power), high angle (vulnerability), overhead (pattern, graphic), over-the-shoulder (conversation), point of view (what the character sees), Dutch tilt (unease). Moves: locked-off, push-in, pull-back, pan, tilt, truck, orbit, crane, handheld.
A panel template you can copy
Write each panel in this format. It doubles as the prompt for the video stage.
- Shot number and length: "Shot 4, about 5 seconds".
- Size and angle: "Medium close-up, slight low angle".
- Move: "Slow push-in".
- Subject and action: who is in frame and what they do.
- Setting and light: where, what time, light direction.
- Sound: dialogue line in quotes, effects, music cue.
- Transition: how it cuts to the next shot.
Before and after: storyboard versus straight to video
Before: a creator types a one-line idea, "A day in the life of a florist", into a video model eight times, getting eight unrelated clips with eight different florists, different shops and different light. Half the credits go on clips that cannot be cut together.
After: the creator writes a six-panel board first.
- Panel 1, wide, locked-off, 4 s: a small flower shop at dawn, shutters half up, a woman in her thirties with a green apron unlocks the door. Soft blue morning light.
- Panel 2, close-up, push-in, 3 s: her hands trimming rose stems on a wooden bench, petals falling.
- Panel 3, medium, truck right, 5 s: she walks along buckets of tulips, touching the petals. Warm light now through the window.
- Panel 4, extreme close-up, locked-off, 3 s: water drops on a peony.
- Panel 5, medium close-up, slight low angle, 5 s: she hands a bouquet across the counter and smiles.
- Panel 6, wide, pull-back, 6 s: the shop from outside at dusk, lights on, flowers in the window.
The stills are generated with the same character references and the same grade note. Two are redone at the image stage because the light was on the wrong side. Then all six are animated. Every clip cuts with the others because they were designed together.
Step by step
- Write the idea in one sentence and the ending in one sentence.
- List the shots in order, using the panel template.
- Vary shot sizes: open wide, move closer as emotion builds, use inserts to cut away.
- Design the characters first so every panel uses the same references.
- Generate the stills. Check the sequence as a slideshow. Does it tell the story with no motion at all?
- Fix panels that break light direction, screen direction or wardrobe.
- Animate each approved still with a motion-only prompt.
- Cut, add sound and grade.
ClickPop's Create Video has a storyboard mode that draws the stills first, so you can approve each frame before anything moves. For longer pieces, Productions turns one brief into a cast, a shot list and a finished cut, and Drama Studio produces a storyboard for every episode of a vertical series.
Boarding a dialogue scene
Conversations have their own pattern, and it is worth boarding them on purpose because AI handles them best one speaker at a time.
- Start with a two-shot that shows both people and where they sit. This sets the geography.
- Cut to over-the-shoulder shots for each speaker, from opposite sides of an imaginary line between them. Keep the camera on one side of that line so the characters keep looking at each other.
- Move to close-ups as the conversation gets more emotional.
- Add reaction shots: the listener's face while the other line plays. These are also your safety net if a take's lip sync looks off.
- Put each line of dialogue in quotes in its panel, with the speaker named.
A simple six-panel dialogue board: two-shot, over-the-shoulder on speaker A, over-the-shoulder on speaker B, close-up on A, reaction close-up on B, return to the two-shot as the scene ends.
Mistakes to avoid
- Boarding only the pretty shots. Board the transitions and inserts too.
- Every panel the same size. A board of medium shots becomes a flat video.
- Skipping the slideshow test. If the stills do not tell the story, motion will not save it.
- Writing motion into the image stage. Stills should be strong frames; save movement for the video prompt.
Questions
Do I need a storyboard for a short AI video?
For a single clip, no. For anything with more than two or three shots, yes. It is the cheapest way to make the shots match and to avoid paying for clips you cannot use.
What should a storyboard panel include?
Shot number and length, shot size and angle, camera move, the action, setting and light, sound, and how it cuts to the next shot.
How do I turn a storyboard into an AI video?
Generate each panel as a still with consistent characters and light, approve them as a sequence, then animate each still with a prompt that describes only the motion.