Questions we hear before the pilot.
Didn't find your answer — talk to sales, we'll reply by email.
How realistic is the result?
The background, lighting, and contact shadow are recalculated for the new scene; the body itself is never repainted — color, chips, and scuffs stay exactly as they were in the source photo.
What about the license plate?
You specify the plate — a dealer plate, blurred, or anything else. The system fits it into the photo's perspective and lighting.
What kind of source photos work?
Any: phone shots, taken outdoors, in the rain, with other cars in the photo. We need at least 1200 px on the long side to deliver a 2048 px result.
Can we use our own backgrounds?
Yes. Upload a background to the library — via the API, the dashboard, or at install time.
How are API tokens counted?
One processed photo equals one token, priced to your volume. Larger batches cost less per photo.
Why is on-premise priced on request?
It's a license plus rollout and support, and the price depends on volume and server count. We'll send a quote after a short call.
Is there a free trial?
Yes: send thirty of your own photos and we'll process them for you.
What hardware do we need for an install?
The system runs on our servers or on yours — about 600 frames an hour. We size the configuration to your volume.
How long does rollout take?
The dashboard works as soon as you sign up. For the API or an install on your own servers, we'll scope it to your setup.
How does the system recognize the car?
Our own DINOv3 Car ID model, trained on millions of photos: make, model, body type, and five dimensions in millimeters from a single photo. More on the Technology page.
Where do the photos go?
We never use your photos for training. In our cloud (API and dashboard) they're stored encrypted, with retention terms set in your contract. With an on-premise install, they stay on your servers — nowhere else.