The AutoLensAI film
One guided walkaround on an iPhone becomes a 360 spin, listing photos and social posts, with your car placed in a rendered studio and its own pixels left untouched.
Film by Daniel Anthony Romitelli Jr. Independent AutoLensAI concept; Chevrolet, Camaro and Z28 are trademarks of General Motors, and no endorsement is claimed. Credits
Intelligent ground-truth capture
The app walks the photographer around the car one stop at a time, 24 stops 15 degrees apart. A cyan lane on the ground marks where to stand next and turns gold once that stop is taken, and each photo takes itself when the phone is steady at its stop.
Voice-guided AI agent
Greets the car, then answers questions while you walk.
Back-office vision agent
After uploadReads every photo after the upload: dust, water spots, condition to preserve, and what it cannot tell from the picture.
Findings guide later treatment
24-stop orbital sampling
01/24
Stops 15 degrees apart
AI agent dispatched • 2022 Ford F-150 identified
Auto-capture
Hold the car in frame. When the phone is steady at a stop, it takes the photo itself.
Details customers notice
A basic cutout can look fine until it reaches an antenna, mirror, wheel, or roofline. CarSegNet checks the whole vehicle, then NeuralSegJet gives those edges a closer pass before the finished frame moves forward.
The edge pass protects antennas, rooflines, and mirror stalks that a rough cutout can erase.
Spokes and open space stay separate, so the new setting shows through where it should.
The same boundary rules hold from one angle to the next: 99.08% mean vehicle overlap across 608 held-out frames.
Drag to compare · or focus and use ← →
Ready to post
2022 Ford Mustang Shelby GT500 Heritage Edition, recomposed for every canvas your team posts into. Each one is its own render from the same photograph. The vehicle is whole in all four, never a crop that lops off a bumper to fit a square.
1:1
Feed post
4:5
Portrait feed
9:16
Stories and Reels
16:9
Website and YouTube
And the caption, written for you
The 2022 Shelby GT500 Heritage Edition in Brittany Blue with white Le Mans stripes. One photograph, composed for every placement you post into. Same car, same paint, same stripes, same wheels. #Buffalo #WNY #716 #ShelbyGT500 #Mustang #carsofinstagram
Every card is its own render from the same source frame rather than a crop of one. Captions are drafted for review before anything goes out.
The scene library
The same photograph of the same car, placed in four different rooms. Nothing about the vehicle changes between them. The studio, the plaza, the showroom and the warehouse are all built around one frame.
Demonstration vehicle courtesy of WNY Dream Cars, shown with permission. The vehicle is their unmodified photograph; every room behind it is generated and labelled as such.
Real cars, real listings
Original photo, then the finished listing. Tap either photo to look closer.
Lot 01 / 12
Lot 02 / 12
Lot 03 / 12
Lot 04 / 12
Lot 05 / 12
Lot 06 / 12
Lot 07 / 12
Lot 08 / 12
Lot 09 / 12
Lot 10 / 12
Lot 11 / 12
Lot 12 / 12
Source photography is credited on every vehicle. Individual preparation notes appear with frames whose treatment differs.
Walk around the car and the app guides you through 24 stops, taking the photos as you go. Driveway, parking lot or bad light, it does not matter.
Every frame is upscaled, the background is cut and replaced with a showroom-grade scene, and the description writes itself.
The finished set comes back to you for review. Nothing is published until you approve it. We build custom integrations with your inventory and dealer management systems.
The engine
Anyone can ask an image model for a showroom. The product is what guarantees the car in that showroom is your car, and that guarantee comes from a segmentation engine we built, measure, and keep training.
A pipeline like this is only as good as its segmentation, and generic models fail exactly where cars are hard: glass, chrome, wheel gaps, the boundary where a halo forms. Ours is fine-tuned on months of human-labelled vehicle data and holds under locked evaluation gates. We proved it in both directions on our own test truck. With our mask, the vehicle survives byte-for-byte; without it, the render came back wearing another maker’s badge.
Where we composite, your vehicle’s own pixels are placed back into the generated scene and the result is measured: on our verification frames, pixel drift across the vehicle interior is 0.000. A frame that cannot pass its check is refused or delivered clearly labelled, never quietly published. The check ships with every frame, not in a slide.
The input standard is a phone photo. Gravel driveway, dappled shade, whatever the lot gives you. Frames for a vehicle render in parallel and land the same day the car hits the lot.
Every job records its full provenance: source, mask, scene, result, and the checks it passed. With your permission, any frame you flag becomes a labelled training example, pixels attached, with zero re-diagnosis. The system that serves you is the system your corrections are training.
In the lab: reconstructing genuinely clear glass from the frames of a walkaround, and physically-grounded relighting driven by measured studio illumination. The interior is stable across views while reflections move, so the reflection can be separated out of your own pixels rather than painted over. The engineering write-up matches the shipped code, file and line.
For dealerships
A car that arrives on Tuesday and goes online on Friday spent three days depreciating in a queue. The same walkaround, shot on a phone and run through AutoLensAI, comes back retouched on a consistent backdrop and ready for your approval. Every unit looks like it was shot in the same studio, because it effectively was.
photos per vehicle
A full walkaround, interior, engine bay, wheels, and every flaw a buyer will ask about.
intake
Trade-ins and auction buys arrive faster than a photographer can get to them.
to edit it all
Cutting backgrounds, correcting exposure, matching a house style, then uploading the set.
The delivery wall
One car, composed four times. No bumper gets cropped to make a square, and nothing posts until a person approves it.
Demonstration vehicle courtesy of WNY Dream Cars, shown with permission. The vehicle is their unmodified photograph; every canvas behind it is generated and labelled as such, and every outward-facing frame still requires review.
