How to Write Video Prompts That Sell
The model already sees your image. Your prompt's job is to tell it what should happen — and that's where most people leave performance on the table.
Describe motion, not the noun
"Product on a table" gets you a static clip. "Slow dolly-in as morning light rakes across the label, hands lift it into frame" gets you an ad. Direct the camera move, the lighting, the action and the pacing.
Match the mode
- UGC — describe the creator's action and delivery; the persona controls who's on camera. Guide
- Single Image — describe the camera move and motion; no person. Guide
- Transition — describe the motion that bridges frame A into frame B. Guide
Ground it with product context
Open Product context and add a product name or store URL so the AI works from your real copy, not a guess from pixels. It's optional but noticeably improves accuracy.
Stay under 2000 characters
The model hard-caps prompts at 2000 characters; the on-page counter keeps that budget visible and turns red at the limit. Spoken-language directives count against it too, so add the language chip rather than padding the prompt.
Let AI draft, then you direct
Write with AI produces a full, duration-paced brief from your image and brand voice. Edit it like a director's note. For the fuller playbook see Writing prompts that work.
Open the generator and write your first brief.
More from the blog
Choosing an Image-to-Video Model for Product Ads: What Actually Ships
A grounded look at what separates a good demo reel from a converting ad — on-model fidelity, direction, language, and workflow fit.
Native-Language Ads: 10 Markets From One Photo
Localised creative beats translated captions. How to ship lip-synced ads in 10 languages from a single product image.
AI Video Ads for Dropshipping: A Practical Playbook
Dropshipping lives or dies on creative volume and speed. How to spin up testable video ads without touching the product.
