Storyboarding Multi-Scene Stories
Turn 8-second clips into real stories: shot lists, cross-scene consistency tricks, and VAR2 workflows that chain it all together.
A 30-second ad is not one generation — it is four to six shots, planned on paper first. Pixar storyboards every film before a single frame is rendered, and AI filmmaking works exactly the same way, just faster. This lesson turns you from someone who generates clips into someone who ships stories.
Think in Shots: The Shot List
Before generating anything, write a shot list: one row per shot, with description, camera move (from the camera language lesson) and duration. A tight micro-story follows a three-act rhythm even in 30 seconds: setup, complication, payoff. Here is a shot list for a coffee brand spot:
| Shot | Description | Camera | Length |
|---|---|---|---|
| 1 | Alarm clock at 5:59, dark bedroom | Static close-up | 4s |
| 2 | Barista Maya unlocks the cafe, flips lights on | Handheld follow | 8s |
| 3 | Espresso pours, steam rises | Slow dolly-in | 6s |
| 4 | Maya hands the cup to the first customer, smiles | Medium static | 8s |
| 5 | Logo cup on the counter, morning sun flare | Slow orbit | 4s |
The Consistency Problem — and Its Fixes
Video models have zero memory between generations. Shot 2's Maya will be a different woman in shot 4 unless you force consistency. Four techniques, in order of power:
- Character passport: write one detailed description paragraph (age, hair, wardrobe, one distinctive detail) and paste it verbatim into every shot's prompt — never paraphrase
- Stills first: generate a consistent set of scene stills using reference images (see the reference images lesson), approve them, then animate each with image-to-video
- Constant style words: keep the same lighting and palette phrases in every prompt — warm golden hour, muted earth tones — so shots cut together naturally
- One model per project: Sora 2 and Veo 3.1 have different looks; mixing them mid-story reads as a glitch
Shot 3 from the list above, passport included
Slow dolly-in on an espresso machine as rich dark espresso pours into a white ceramic cup, steam rising and catching the light. In the soft-focus background, Maya, 34, curly dark-brown bob, small silver hoop earrings, mustard-yellow apron over white tee, works at the counter. Warm golden hour light, muted earth tones, cozy specialty cafe.
Even a background appearance gets the full passport — that is what keeps shot 3's Maya matching shot 2 and shot 4.
Chaining It with VAR2 Workflows
VAR2's workflow canvas is built for exactly this pipeline: chain image nodes into video nodes, so each scene still feeds its own image-to-video step, and run the whole storyboard as one connected flow. Change the character still once and regenerate downstream shots — instead of babysitting five separate generations.
Build your five-shot storyboard as a connected canvas — stills flowing into videos. Open the workflow canvas