Notlith · Burlington
AI video production in Burlington, Vermont: a launch spot for a small creamery’s new pint
Imagine a small ice cream maker in Burlington, a city where a famous pint brand began in a renovated gas station in 1978. It is launching a new flavor in grocery freezers across Vermont and wants one commercial for social feeds, store screens and its website. This page follows how Notlith would build that spot with AI from the creamery’s own photographs, and where the AI has to stop.
Remote production · From your photos


Count the pieces before animating them
A new flavor is sold on what is inside: chunks of brownie, a ribbon of caramel, maple pieces. An image model asked for a scoop will add swirls and chunks freely, and a customer who opens the pint at home will notice. That gap is the fastest way to lose a repeat buyer.
We start from photographs of a real scoop cut from a production pint, side by side with a pint opened from the top. Generated motion can turn the scoop, frost the edges and let caramel catch the light, but the amount of inclusions matches the photograph. Each frame is compared with it before editing.

The word Vermont on the lid
Vermont regulates claims about Vermont origin, and the Attorney General has issued guidance for businesses using the state’s name on labels. If the lid says the cream comes from Vermont dairies, the spot can say it; if it does not, the film adds nothing.
We copy origin and ingredient wording from the approved label and keep it on screen as text, not as a looser spoken line. Pastures, barns and cows are not generated to suggest a farm the creamery does not own or buy from. If the creamery has real photos from its dairy supplier, we use them with credit.

A melt that behaves like ice cream
Close-ups of melting are where generated video often fails, with drips that move upward or a texture closer to cream cheese. The creamery’s own photos, taken a few minutes apart as a scoop softens on a plate, give the model a real reference.
We generate only short moves between those stages and reject any shot where the texture or color drifts from the photographs. Spoons, bowls and the counter come from the creamery’s café or are clearly neutral. No person tastes the ice cream on screen, because a generated face reacting to a product reads as a testimonial.

From the freezer door to the phone
The launch runs in co-op and grocery freezers, on the creamery’s feeds and on a small screen by its scoop counter. Each place sees the spot at a different distance: a thumbnail, a door at arm’s length, a screen across a room.
We deliver a thirty-second master, a vertical version that opens on the pint lid, and a short silent edit for the counter screen with the flavor name large. A note that the scenes were created with AI from photos of the real product goes in the caption. Usage rights and retailer versions are set in the quote.
Fictional concept created with AI
In pictures
Catalog demonstration
See the production approach
English, adaptable to your language
Plan the details
Can you show a flavor before the first batch?
Not as finished product. Until a real pint exists to photograph, we can only show a clearly labeled concept.
Can the spot mention that Ben & Jerry's started in Burlington?
As history, perhaps, with legal review. We keep the spot about your own pint and do not suggest any link to another brand.
Will the AI change our label design?
No. The label is placed from your artwork files and checked frame by frame.
What should the creamery send first?
Photos of the pint from every side, a scoop cut from a real pint, melt photos, label files and the retailers that need their own end card.
Start with your images
Send us photographs of the pint, the label files and the launch date, and we will reply with a script and a frame plan.









