Thrift and resale listing videos from one flat-lay photo

A reseller can turn a single flat-lay photo of a jacket, bag or sweater into a 6-second listing clip with Wan 3.0 image-to-video, and at 480p the clip costs $0.375, charged as 38 cents. At 720p it is $0.75.

Listing video helps buyers who cannot touch the item, but resale has a special risk: a generated clip must not make a worn item look new. This post sets the rules, then the cost for a batch of 50 items.

What first-frame video does

With image-to-video the photo becomes the first frame and the model adds motion, such as a slow camera push or fabric catching light. The Video Generation docs list Wan 3.0 at 2 to 30 seconds with end-frame and reference support, and aspect ratios from 9:16 to 16:9. For first-frame use, the existing post on first frame, end frame and Grok limits is the detailed map.

The photo must be a public HTTPS link. Upload your flat-lay to your own storage or your Sume workspace and pass the URL.

Rules for honest resale clips

Generated motion is decoration, not evidence. Buyers decide on condition, so the listing photos of flaws, tags and measurements stay as photographs, and the clip should never smooth over a stain or hide a worn cuff. Many marketplaces have their own rules about how media must represent an item, and Sume does not know them, so read the marketplace's current policy page before you post video there.

For clothing, prompt for camera motion over the garment rather than invented movement of the fabric. A slow push-in and a light sweep keep the shape of the item. Avoid prompts that add a person, since a generated model wearing the item implies a fit you have not shown.

• Keep flaw photos as photos, next to the clip.
• Prompt camera motion, not fabric animation.
• Do not add a generated wearer unless you label it.

Cost for a 50-item drop

Sume prices Wan 3.0 at the provider list times 1.25: $0.0625 per second at 480p, $0.125 at 720p, $0.25 at 1080p (read 2026-10-10). Fifty 6-second clips at 480p are 50 x $0.38 = $19.00. At 720p they are 50 x $0.75 = $37.50.

Running the batch

Fifty items is a job for a bulk run rather than a loop in a notebook. Sume's bulk runs take 1 to 100 items with concurrency from 1 to 16 and an idempotency key. A queue marked completed does not mean every item succeeded, so read each item. The post on completed is not succeeded covers the requeue step.

There is also a public format called sume-virtual-try-on under the handle sume. Read GET /v1/formats/sume/sume-virtual-try-on to see its inputs before you rely on it, and label any output that shows a generated wearer.

Choosing a source photo that survives motion

The best source is a sharp, evenly lit, top-down flat-lay with the whole item in frame and a plain surface behind it. Busy backgrounds invite the model to invent details, and a cropped hem or sleeve gives it room to guess. Take the photo with the item smoothed out and any visible labels facing up, since a clip that keeps the label legible is more useful to a buyer than a pretty one.

Shoot each item the same way so that one prompt template works for the whole batch. A short prompt such as "slow camera push toward the center, soft light sweep, nothing else moves" is repeatable and leaves little for the model to improvise. Test it on three items of different color and texture before you launch the batch, because dark knits and shiny leather behave differently from pale cotton.

Choose your five best-selling categories, run one clip each at 480p ($1.90 total), and compare view and message rates on those listings against your usual photos. Move to 720p only for items where the clip is the main attraction.