Sneaker OEM: the colorway still is the listing
I2V that photo. MJ sneakers are a different SKU.
The listing is the colorway photograph. Image-to-video that still. A Midjourney sneaker is a different SKU. Do not generate a prettier pair and hang your factory's name on it.
AI image to video keeps your upload as frame one. AI product video is the same contract for a pack shot: the object in the ad is the object you photographed. Sneaker resellers already live on trust — authentication, the box, the tells. OEM is stricter. You made the pair. The still is the SKU.
The still is the SKU
A colorway is materials, a last, an outsole, a tongue mark, a lining. The photograph the factory signed is the only file that contains those facts. Text-to-image a "similar" Jordan-shaped object and you get a plausible shoe. Plausible is a different product. Anyone who has held the real pair will read the difference as counterfeit energy even when you meant "lifestyle."
Midjourney is good at sneakers as a vibe. That is the problem. Vibe invents a colorway that does not exist in the warehouse. The overlay tab, the heel counter, the exact grey — those are the listing. They are not a prompt. If you do not have the still, shoot it. Generation is not a photography department for a SKU you already tooled.
I2V the photo you will sell
Image-to-video does not argue about what the shoe looks like. Frame one is already correct. Your prompt is motion and camera: slow orbit, lace settle, light wrap. Re-describing the sneaker is how the model re-invents it.
The still sets the aspect. A 1:1 pack shot animates 1:1. Some models take a first frame and a last frame — lateral still to medial still — so the orbit does not invent the inside of the collar. Both ends are photographs of the same SKU.
Duration is a dial. A five-second orbit of the real pair outperforms a long take in which the overlay starts to crawl. Drift is the model running out of information about the side it has never seen. Shorten the clip, ask for less motion, or pin both ends.
MJ sneakers are a different SKU
Do not text-to-image a "better" colorway and I2V that. Do not mix a generated pair into a carousel of factory stills. Do not "clean up" the overlay tab with a fashion model. Those files are new objects. They will not match the box or the pair that lands on a doorstep. Reseller trust content exists because buyers already assume the photo might be a lie. OEM generating a prettier shoe is how you become the lie.
If you need a scene the warehouse cannot shoot, that is still a reference job. Photographs of the real object stay the ground truth. The mountain is allowed to be invented. The colorway is not.
What the listing cut actually is
One locked still of the colorway. Optional short I2V of that still. Optional first-and-last orbit between two factory angles. Type for the SKU name, burned as overlay, not generated into the leather. Catalogs sell the pair that exists.
FAQ
Can I generate a colorway we have not sampled yet, as a sales sketch?
Yes, as an internal sketch, labelled as a sketch. No, as the listing for a SKU that is already tooled. The listing still is the sampled pair.
Why not text-to-video a sneaker if the prompt names our materials?
Because a prompt is not the last. The model will draw a fashionable shoe. Image-to-video of the factory still is the only generate that starts on your SKU.
Does a long cinematic take help a drop more than a five-second orbit?
Usually not. Longer I2V is where overlays and outsoles drift. A short orbit of the real pair, then cut. Duration is not a brand flex on a colorway that has to match the box.
What if the only photo is a reseller shot with a busy background?
Clean the still with an edit that keeps the shoe, or reshoot. Do not replace the pair. A generated sneaker on a clean sweep is a different SKU with nicer lighting.