Why should you include a human hand in product photos when generating AI ads?

A human hand in your product reference photos gives the AI model an unambiguous scale anchor. The model has no innate sense of how big your product is — a hand holding the product and each layer of its packaging teaches it true size and proportion, so the product renders at correct scale in every generated shot.

Photograph your product held in a real human hand — the product itself, plus every layer of its packaging, from multiple angles including close-ups — and upload those images as references before you generate anything. This set is what invideo's team calls a product sheet, and the hand is the calibration element: an AI model reading a lone product photo has to guess whether it's palm-sized or shelf-sized, and a wrong guess propagates through every frame of the ad. A hand in frame removes the guess, because the model reads the product's dimensions against a reference object it deeply understands.

Upload the product sheet once into the invideo agent's context so it applies to every subsequent image and video generation in the project — invideo is an agentic video creation tool, and its context tab persists your references across the whole project, so you never re-explain product scale per shot. Use a real photo of your actual hand holding the product rather than asking the model to invent the interaction: the reference carries true grip, contact, and size relationships the model can copy instead of inferring.

The payoff is fewer wasted generations and a product that holds its size across shots. One documented production fed the invideo agent hand-held reference images of a supplement sachet — including loose gummies in the palm — specifically so the model understood the actual size of every packaging layer relative to a human hand; the resulting ads held 100% product consistency, validated across 3 different ads in 2 completely different markets, at roughly $70 per finished localized ad including rejected clips. Without the scale reference, product detail drift gets worse as packaging complexity increases, and each drifted generation is credits spent on a clip you'll discard.

One edge case: if a single ad shows products at meaningfully different sizes — a sachet and a full jar, say — create and lock a separate hand-referenced keyframe for each product scale. A single unified reference invites the model to confuse the two scale relationships mid-generation.

The hand reference solves scale specifically; for overall brand and product fidelity, it works best layered on top of your broader brand context (guidelines, product catalog, visual references) loaded into the invideo agent at project start.

Watch some of these to see what works for you:

I give the agent the reference images of the hand holding the sachet, a few gummies in the palm, so it understands the actual size of all of the layers of your packaging compared to the human hand.

— invideo's creative team