Toys: the listing still cannot morph into a different toy
Identity hold. Omni will invent features.
Identity hold. Omni will invent features. A toy listing dies when the figure in motion is not the figure in the box. Joint colour, printed face, the accessory that is actually in the set — those are the product. Text-to-video will redraw them. A multi-shot storyboard will restage them. Image-to-video is how they stay. Animate the listing photo. Do not prompt "our 12-inch articulated hero" and hope the model remembers the hinge.
The SKU cannot become a different toy because you typed "deluxe."
The SKU is already in the still
If you would not print the still on a box, do not buy motion. Fix the photo. Even light, the toy large in frame, face and print readable. That still is cheaper to argue about than a 20-second clip with the wrong visor.
Seedance 2.5 Image to Video is the still-to-motion row: one locked frame into a continuous clip, up to 30 seconds, up to 720p, native audio with the picture. A second image sets the closing frame when the shot has a destination — front of pack to three-quarter of the stand. Image required. Forty-seven credits listed. Features: image-to-video, first frame, last frame, long form, audio sync.
Seedance I2V starts there. Camera and a slow orbit are the prompt. Identity is already in the pixels. Repeating "exact SKU, do not change the figure" in text is how you invite a renegotiation.
Thirty seconds in one pass is the production reason this family belongs on a product-page loop. Stitching four short text-to-video clips around a toy is how the print drifts by second twelve. 720p is the ceiling. Native audio is on — write the mix, or you will get a band you did not ask for. Last frame is optional: use it when the second angle is a second approved photo. Do not attach a last frame of a different SKU. That is how figure A closes as figure B.
Lock the still before you spend on motion is the credit argument. Video meters are a terrible identity debugger.
Omni restages. Restaging is a new toy.
Kling 3 Omni multi-shot is not Movie Mode. Omni takes one prompt and returns up to six distinct shots inside a single generation, with duration, angle and camera move specified per shot. The usual window is a 15-second generate. You still pay for seconds. Identities, grade, and world have to survive inside that generate. There is no scene planner, no chain off the last frame of shot four, no combined export of three approved stills. You get one file with cuts in it.
Useful for a three-beat gag. Not useful for a hero pack shot that must match a still. Multi-shot restages. Restaging is drift. Drift on a toy is a new feature: an extra joint, a different visor, a third accessory that is not in the box.
Do not quote an Omni clip as the listing video. Coverage is not identity. Omni will invent features because a six-shot generate has to fill six cameras, and filling cameras is how a figure grows a second pair of hands. Stitching Seedance lock plus a cutaway in an editor is a different object. Omni is not that stitch.
Pause on the hinge
Run the I2V take. Pause. Read the face print, the joint colour, the accessory count. If any of those would fail a listing audit, throw the clip away. Do not "fix" it by regenerating Omni or text-to-video. Make a new still.
If they match, you have a product move. Caption after. A new SKU — a different mould, a different face print — is a new still, not a prompt adjective. The listing still cannot morph into a different toy.
FAQ
Can I text-to-video a toy if I describe the listing carefully?
You can. You will re-cast the figure every take. Face print, joints, and accessories are the expensive failure. I2V the listing photo so those are not sampled again.
Why not use Omni if I need several angles of the same toy?
Because Omni restages inside one short generate. Several angles of a SKU are several approved stills, then I2V, then a stitch — or first-last when the shot has a destination. Six cameras in fifteen seconds is coverage. It is not a pack-shot department.
Is 720p enough for a PDP loop?
For feed and most product-page loops, yes. For a 2K plate headed for a grade, pick a higher-resolution generate and still start from a locked still. Resolution is delivery. Identity is I2V.
Should the playset in the background be generated?
Only if the still already has the playset you will ship, or you accept that the room is atmosphere. If the accessory is in the box, lock it in the photo. Do not let motion invent a second set.