Tint is a glass still; do not let the model invent the car
Approved vehicle photo in. I2V out.
Approved vehicle photo in. I2V out. Window tint is sold as glass on a specific body: this sedan, this windshield cut, this visible-light transmission. A generated SUV with plausible dark glass is a cousin. It is not the car in the bay, and it is not a clip you can put on the shop page.
Photograph the vehicle you tinted, or the vehicle you will tint. Animate that photograph. Do not animate a car the shop never had.
The glass is already in the photograph
A tint quote is a body, a film, and a set of windows that have to stay the windows on that body. Text-to-video does not know the A-pillar. It will invent a stock crossover with melted lettering on the plate and a greenhouse that is not on the work order.
Image-to-video starts from the file you upload. Frame one is the still. Body lines, existing paint, the shop floor, the film that is already on the glass — those survive because they were already pixels. The prompt is camera and physics: a slow orbit, a light rake across the windshield, a hold on the side glass. Re-describing "a tinted luxury sedan at golden hour" is how the model starts drawing a second sedan.
Seedance 2.5 is the family page; the image-to-video row is the still-to-motion job for that plate: first frame required, four to thirty seconds in one pass, native audio, 720p ceiling, optional last frame when you have a second approved photo — untinted in, tinted out, same angle, same vehicle. A last frame is control. It is not a default. If you only have the finished job, one still and a slow rake across the glass is the job.
If the still would not go in the job folder, do not buy seconds. Crop the hoist. Do not prompt it out and hope the model rebuilds the quarter panel.
Text-to-video will pick a different lot
Text-to-video has no plate. You asked for "a car with ceramic tint at dusk," and the model decides whose car. Average is not your VIN. Average is a greenhouse that does not match the vehicle in the bay.
That is lock the still before you spend on motion, except the miss sells the wrong body. Do not prompt the make into a text-to-video row and call the missing jpeg a technicality. A sentence is not a windshield.
Seedance I2V is still the still-to-motion job is the router. This page is only the glass. Thirty seconds in one pass holds a body; stitching four short text-to-video clips is how the greenhouse drifts. 720p is the ceiling on this endpoint.
Need the same car on a lot you did not shoot? That is the AI product video generator: reference-to-video off photographs of the actual vehicle. I2V from the bay still. R2V when the lot changes and the car must not. T2V for neither.
Motion is a rake, not a second body
Once the vehicle still is locked, the generate may move the camera and land on a second approved still of the same car. Two photos of two different cars is a morph of the quote. It may not add a second row of windows or invent a ceramic badge on the glass. Keep the greenhouse large in frame.
Prompt only the move. Leave make, model, and the film name out of the paragraph; they are already in the still. Repeating them is how Seedance starts renegotiating the lettering on the plate. Pause on the windshield. Is it the car in the quote? Yes: ship. No: you used text-to-video, or you first-last'd two bodies.
FAQ
Can I prompt the make and the tint percentage into text-to-video instead of uploading the bay photo?
No. A sentence is not a plate. The model will still invent the greenhouse. Upload the approved vehicle photo and describe only the camera.
Do I need a last frame of the untinted car?
Only when both stills are photographs of the same vehicle, same camera height, and the shot is meant to travel between them. A single finished-tint hero with a slow rake across the glass is the usual clip.
Why not generate a hero car if the shop floor is messy?
Because the messy floor is cheaper to crop than a invented body is to explain. The photograph is the cheap place to lock body lines. Text-to-video will pick a different lot.
Is 720p enough for a tint shop Reel?
For feed and a Google Business clip, yes. The glass matching the quote matters more than a higher-resolution invention of a different car. If you need a higher ceiling, start from the same locked still — do not switch to text-to-video to chase resolution.