Mobile locksmiths: the van livery has to match the still
If the van is the brand, I2V. A random Kling van is a different company.
If the van is the brand, image-to-video. A random Kling van is a different company. Mobile locksmiths have a phone number and a vehicle, not a storefront. The livery — city, phone, colour the neighbourhood already knows — is the trust signal. Inventing a prettier van in a prompt is inventing a prettier competitor.
Image-to-video is the job when the still is already true. Seedance 2.5 owns that lock on Versely: one approved frame into a continuous clip, 4–30 seconds, up to 720p, native audio, optional last frame. You cannot pin a last frame you never photographed.
The van is the identity lock
Wrong door graphic in a "local" ad is a lie the caller notices in the driveway. Not a styling miss. A different fleet.
Park the actual van. Even light. Livery readable. Phone number in the plate, not in the prompt. Describe only motion: orbit, rain, a pull-back from lock to door. Re-describing the company is how the model re-typesets the side panel. The still-to-motion job does not care if the object is a bottle or a van. Identity locked. Camera allowed. A drifted phone number is a failed take.
A generated van is a different company
Kling 3 is a motion engine. It is not a licence to skip the photograph. Text-to-video of "a white locksmith van, 24/7 on the door" will give you a white van with some 24/7. Almost right is a different company — the same almost fake operators already use in search.
If you have the still, you are in I2V. Run the still. A three-beat lock gag is a second shot, not a reason to throw the livery back into a prompt. The proof is the van they will see, not a talking generated locksmith.
Photograph once, reuse the asset
The van does not change every week. The callouts do: locked out, rekey, after-hours.
Set up reusable characters and products is the named job. The agent opens an Asset Builder card for the van — and several cards at once for a logo, a lock, a storefront. Saved assets are referenced by key across scenes. You do not re-upload the livery, and you do not hope two text-to-video rolls spelled the phone the same way. Uploading references does not spend generation credits. Animating the still does.
A last frame is optional: three-quarter to door graphic, both photographs of the same vehicle. Two generated "ends" are two guesses.
The 100% pause
Before you run ads, pause the clip on the side panel. Read the number. Read the city. Read the mark you are legally allowed to show. If any of that is the model's handwriting, you do not have a locksmith ad. You have a clip of a van that does not exist.
Then ship the callout as type or as a caption on the true plate. The vehicle stays the vehicle. The week's message sits on top of it. That is how a one-truck operator looks like the same company on Tuesday and on Saturday without rolling a new fleet.
FAQ
Can I text-to-video a van if I describe the livery in detail?
No. Detail in a prompt is not a photograph of the livery. The model will still invent type, colour, and layout. Photograph the van, then animate the still.
Is Seedance 2.5 the only model that can move a van still?
No. Image-to-video is the tool shape; Seedance 2.5 is the identity-lock row — 4–30 seconds, 720p, native audio, optional last frame. Other I2V rows exist. Start from the still, not from "a van."
Why upload the van as a reusable asset if I already have the photo?
Because next week's rekey ad will need the same vehicle. A saved asset is a key the workflow can reuse. A one-off upload in a one-off generate is how the third clip quietly becomes a different van.
What if I do not have a branded van yet?
Then you do not have a van still to lock. Film a lock, a door, a licensed person you have the right to show. Do not generate a fleet you do not operate. The fake-van problem in this category is already the buyer's entire suspicion.