Videos
AI product video maker
If you sell a physical thing you already have the raw material: the photographs from the listing. The work is deciding which shot opens, which ones group together, which scene can hold four of them at once, and how each picture sits in its box without being cropped through the subject.
The photos are read before anything is designed
Every uploaded image goes through a vision pass that returns what it shows, a few key points and whether it is usable as a scene asset. Separately, the real pixel dimensions are read out of the file header rather than guessed by the model, because a guess is wrong often enough to miscast a slot. Images that clearly belong together are grouped, and the group is shown in one scene in the order given.
Shape beats resolution
Every image slot publishes the box it makes in each canvas, and every uploaded picture is tagged wide, square or tall. The director matches one to the other, so a tall packshot lands in a tall slot and a wide lifestyle shot in a wide one. A mismatch does not fail, it crops or letterboxes, which is the single most common reason a product video looks careless. Uploaded pictures are also never upscaled, except a hero slot that fills at least 60 percent of the frame, and even then by no more than two and a half times.
Scenes that hold more than one
Gallery is the workhorse and its layouts hold two to eight pictures: three tiles in a row, three tall film frames, a strip the camera pans across, four dealt cards, a tilted full bleed wall implying a big catalogue, cards standing on a perspective floor, and two that walk four images one at a time into a pile or a grid. Showcase is for a single hero: one shot in a glass card, held centre frame, tilted diagonally, or full bleed and panning.
Frames make each photo its own beat
A scene built from a container of several picture and caption pairs is a frame sequence, which means each photo enters with its own caption, holds, then gives way to the next. The timeline draws one drag handle per frame, so you can give the hero shot three seconds and the detail shots one each. You can also write a separate narration line per frame, so the voice describes the picture actually on screen.
How it works, in three steps
Step 1
Upload the shots you would put in a listing
The hero, two or three details, one in use, one showing scale. Eight images is the cap, and the weak ones get dropped anyway.
Step 2
Say what each thing is for
The brief is passed to the vision pass as context, so telling it what the product is makes the image descriptions far more accurate.
Step 3
Retime the gallery
Drag the per frame handles in the timeline so the shot that sells gets held and the supporting ones move quickly.
The full walkthrough with screenshots is in the guide Make a promo video with AI voiceover.
Limits worth knowing
- Eight images per generation. Each file may be up to 15 MB, in PNG, JPG, WebP or GIF.
- Backgrounds are not removed and photos are not retouched. A messy shot stays a messy shot.
- There is no 3D turntable, no product spin and no animated packshot. Photos are moved and framed, not rebuilt.
- Generated images are only used where a layout wants a generic visual. The engine will not fake your product.
Questions people ask
How many photos do I need?
Four is usually enough for a short video and six gives a standard one room to breathe. More than eight cannot be uploaded, and beyond about six the director starts choosing a high capacity gallery layout rather than giving each photo a beat.
Will it crop my product out of frame?
It tries hard not to. Each picture is tagged wide, square or tall and matched to a slot of the same shape, and you can override that per image in the inspector with the shape toggle.
Can I add a price?
Yes, in the brief. A price becomes an offer scene or a stat, both of which set it in large type. It will not invent a price, so if you do not write one, none appears.
What about a video clip of the product?
Upload one and a clip scene plays it with trim controls and a fit setting. The clip plays muted by default, which is normally what you want under a voiceover.
Make your own video
The button opens the generator with this use case already described. Change the wording to match yours, generate, then edit anything you like.
Create a video with OneCraftRelated pages
Ecommerce product video maker
A product video for a marketplace is written against a rulebook, not a brief. The big retail platforms autoplay it muted, cap it at well under a minute, and ban a surprising number of the things a marketer would naturally put in it. Knowing the bans before you write is most of the job.
AI images for video scenes
Some scene layouts want a photograph and you do not have one. A generated image fills that gap, and the interesting part is the rule around it: it is only allowed where the picture is genuinely generic. Faking your own interface, your logo or a real person is explicitly forbidden.
Turn product photos into a video ad
A listing photo set is a good half of a video ad and a bad whole one. The pictures answer what it looks like. An ad also has to answer why now, how much, and what to do next, and none of those three are anywhere in a photograph, so they have to come from you.
More finished work of this kind is on the video examples hub.