Walmart marketplace stills are compliance photos first

Do not MJ the pack. Lock the still, then I2V. Words on the label are Ideogram or overlay, not a prompt. A Walmart marketplace listing is a compliance photo of a SKU. It is not a cinematic pack shot, and it is not a place to discover whether the bottle is yours.

The still is the contract. Motion is optional. Lettering is a typesetting job.

The listing still is the SKU

A marketplace thumbnail has one job: the buyer should recognise the unit they will receive. Angle, colourway, pack size, and the words on the label are identity. If any of those drift, you did not make a "more premium" image. You made a different product.

Midjourney is a look. A listing is a match. Iterate in stills until someone can point at a frame and say "that is the pack we ship." Lock the still before you spend on motion is the routing rule. This post is the marketplace version: the still is the listing.

Do not Midjourney the pack

Cinematic lighting, a three-quarter orbit, a kitchen that is not your kitchen — those are ad frames. They are allowed after the SKU is signed, and only as image-to-video off the signed file. They are not how you invent the pack.

Text-to-video is worse. It invents the bottle and moves it. When the label comes back almost-right, you paid video credits to fail at identity. Same prompt, new noise, new pack.

AI image-to-video treats the upload as frame one. The prompt describes camera and action only. That is the honest way to put a listing still in motion without renegotiating the SKU.

Label words are Ideogram or overlay

Do not put the legal name, net weight, or flavour line in a video prompt. In-frame lettering has to be right on every frame. A still has to be right once. On-screen text in generated video is the typeset-don't-prompt rule.

Ideogram V4 is the catalog row for that still: text rendering, posters and logos, six credits flat. Run it from text-to-image. Approve the letters. Then I2V that file if you need motion. A turning label can still drift — that is the standard worst case — but you start from a correct frame one.

If the words are a headline or CTA that nobody prints on the pack, do not generate them. Add a text overlay burns one authored line — top, center, or bottom — from the font registry. Pixel-identical every frame. Change the copy without regenerating the product.

• Words on the SKU. Ideogram (or another stills row that renders type). Lock the still.
• Words you wrote for the ad. Overlay. Never a prompt.

Motion after the SKU is signed

If this still were printed on a box, would we ship the box? If no, do not buy seconds of motion. If yes, write a motion line that names only movement. Leave the flavour name out of the paragraph. Identity is already in the pixels.

A regen of I2V on a locked still is a new take of the same listing. A regen of a cinematic paragraph is a new casting session. Only one of those can match the pack on the shelf.

FAQ

Can I generate a "lifestyle" listing and crop it for the marketplace?

Not as the compliance still. A lifestyle frame restages the SKU. Cropping a restaged bottle does not make it the pack. Shoot or generate the listing still as the listing still. Lifestyle is a second asset, off the same locked product, if you need it.

Why not prompt the flavour name into Kling or Veo?

Because a video model draws the word again every frame. One wrong letter on a flavour line is a listing problem, not a style problem. Typeset the label in a still, or overlay the line. Do not make the generate responsible for regulated words.

Is overlay enough for a nutrition panel?

No. Overlay burns one fixed line for the whole clip. A panel is a dense layout. That is a stills job — Ideogram or a real photo of the pack — not add_video_captions. Use overlay for the hook, the badge, the CTA.

Do I need video at all for a Walmart SKU?

Often no. The listing is a photo. Buy motion when you have a reel, a PDP loop, or a paid placement that actually needs seconds. Until then the locked still is the deliverable.