Step 4: Concept & Script
Add scenes with visual descriptions, dialogue, and speaker assignments.
What is the Concept & Script step?
The Concept & Script step is where you define what your video will say and show. On Multi-Shot and Product Marketing layouts, add scenes with visual descriptions and dialogue, assign speakers, and monitor the estimated duration against your duration target from Step 1. On Hyper-Realistic and Custom layouts, you get a single Scene Description field — write what the viewer sees and any spoken words together there.
Scene Description
The visual setting, action, and framing for each scene. Describes what the viewer sees. On Hyper-Realistic and Custom layouts, this is the only text field — include any spoken words directly in this description, since there's no separate Script/Dialogue field on those layouts.
Script / Dialogue
The spoken words for each scene — what the speaker says. Only shown on Multi-Shot and Product Marketing layouts. Hyper-Realistic and Custom layouts don't have this field — write spoken words into Scene Description instead.
Speaker
Name the speaker for each scene (e.g., Host, Narrator). Only shown on Multi-Shot and Product Marketing layouts, where a video can have more than one voice across scenes.
Duration Estimate
Automatically calculated from your script length (and any per-scene duration hints). Only shown on Multi-Shot and Product Marketing layouts. Shown against the duration target from Step 1, with a warning color when you're over.
Configuring Concept & Script
Add scenes
On Multi-Shot and Product Marketing layouts, click 'Add Scene' to create scene cards — each with its own Scene Description, Script/Dialogue, and Speaker fields. On Hyper-Realistic and Custom layouts, you get one simplified editor with a single Scene Description field and no Add Scene button.
Write dialogue
On Multi-Shot and Product Marketing layouts, enter the spoken script in the Script/Dialogue field for each scene. On Hyper-Realistic and Custom layouts, write spoken words directly into Scene Description. Keep it conversational — short sentences work best.
Assign speakers
On Multi-Shot and Product Marketing layouts, set a speaker name per scene (e.g., 'Host'). This field doesn't appear on Hyper-Realistic or Custom layouts.
Review duration
On Multi-Shot and Product Marketing layouts, check the estimated duration meta row — it turns amber when the total exceeds your duration target from Step 1. Hyper-Realistic and Custom layouts don't show this meta row; use the duration control from Step 1 as your guide instead.
Most AI video models cap out at 8-15 seconds per generation, so scripts should be short and punchy — roughly 2-3 short sentences for an 8-second video.
Do
- Write natural, conversational dialogue
- Use short sentences for punchier delivery
- Include a clear call-to-action
- Match script length to your duration target
Don't
- Use industry jargon without context
- Write walls of text without scene breaks
- Write a 60-second script for a 8-second video
- Forget the scene description — visuals matter as much as words