How to make a product video with AI
Upload product photos as references, write a short script with one or two actions and pick the format. Five to ten seconds is enough for an ad: the product appears, does something useful and stays in frame. Reference-based models keep packaging shape better than a text-only prompt.
Updated:
Step by step
Shoot the product
Clean background, even light, the item fully in frame. A hand in the shot is often read as part of the product.
Script with one action
For example: the bottle turns toward the camera and a drop falls. One action per clip gives a clean result.
Format and model
Choose a vertical format for Reels and TikTok. Pick the model from the product video task page.
Where it is in ClickBait
Frequently asked questions
How long should an ad clip be?
Short clips of five to ten seconds often work best in feeds. Each model page shows its supported lengths.
Do I need a voiceover?
No. Sound is a switch on the clip, and on-screen text can be added in editing. A separate voice is made in the Voiceover section.
Why does the product look different in the video?
A strong camera move changes packaging details. Try a locked camera and add a photo from another angle.
How do I pick a model for a product clip?
Reference-based models keep the packaging better. Each model card shows which modes it supports.
Can I make the clip without a product photo?
Yes, from a description. But the packaging shape and color will then be approximate. For a specific product, start from a photo.