10% off monthly plans with code NODE10, until August 31
Drop in the shot you took with your phone: the clutter goes, the light is redone, the number plate stays blank, and what you are selling does not move a pixel. Three listing photos out, workflow visible without an account.
Real workflow, real results. Pan and zoom freely.
Use this workflowThe workflow is right here: the car photo, the instruction that forbids retouching the vehicle, the 4:3 cover, the head on portrait shot for the mobile app, and the object branch that outputs the square packshot. Pan, zoom, open the nodes.
At a glance
On a classified ad, the first photo decides the click, and it is almost always taken in a hurry: the car at the kerb under a grey sky, with the neighbour's bin and a road sign growing out of the roof, or the item sitting on a worktop covered in crumbs. What puts the buyer off is not the vehicle, it is the setting. This workflow takes that photo and changes only what is around it: the background, the light, the framing.
One rule holds the rest together: never retouch what is being sold. Left free, a model erases the scratch, straightens the bumper, descales the machine and repaints the wing. The ad ends up better looking than the object, the buyer sees the gap on arrival and walks away, and a photo that hides a known fault turns against the seller. So the template instruction lists what has to stay visible, and the guide writing a prompt that works explains why that list beats a blanket ban.
On the model side, Flux Kontext edits by instruction: your photo goes into the node and only what the sentence asks for changes, which is exactly what you want here. Seedream 5 Pro and Nano Banana are the alternatives when a part needs to be described more precisely, and the Flux versus Seedream comparison says which one to pick depending on the source photo. The calculator gives the exact cost before you sign up.
Import the photo as it is
No need to shoot it again: the phone picture is enough, even badly framed, even under a grey sky. Pick the one where the item is whole and sharp, because sharpness is the one thing the model cannot invent.
Keep the instruction that protects the item
It names what must not move: the model, the colour, the wheels, the marks on the bumper. Add your own faults to the list, a dent on the door, a worn patch, and remove what does not apply to your item.
Output the cover, then the second view
The 4:3 cover is the photo that sits at the top of the ad. Its output feeds the next node, which renders the same car head on in portrait: the format mobile apps show full screen.
Do the same for an object
The second branch of the canvas starts from an object photo and outputs a square packshot on light grey, the format of classified ad grids. Same instruction, same bans: the limescale and the dent stay visible.
The same canvas, tuned for different needs: each variation is two or three lines of prompt away.
The front three quarter shot, taken at headlight height, is still the best cover. The workflow removes the parked cars behind, the street furniture and the grey sky, straightens the horizon and puts the car on empty tarmac. Keep the signs of use visible: they save you half the viewings that lead nowhere.
A chest of drawers photographed in a lived in living room sells badly, not because it is damaged but because people look at the rest of the room. Ask for a neutral background and soft side light, keep the scratches on the top, and shoot the whole piece rather than a detail: buyers look for proportions first.
This is where the square packshot changes everything: on plain grey, the coffee machine or the console finally fills the whole thumbnail in the grid. Cleaning an object before the photo is fair, making a dent disappear is not, and the template instruction puts those two moves on either side of the line.
A bike leaning against a garage wall disappears into the mess of tools and boxes. Lifted onto a light background, the frame reads, the drivetrain is visible, and the ad stops looking like a junk room. Add to the instruction the parts that must stay as they are: marked saddle, chipped paint, worn tyres.
A professional publishing thirty ads a week cannot rent a studio. The same workflow, applied to every photo, gives a consistent series: same background, same light, same framing, which makes the shop look like a real window display instead of a pile of pictures taken anywhere.
Classified ad apps crop your photo without asking: square in the lists, vertical when opened full screen. The template outputs both formats from the same picture, which saves you from discovering that automatic cropping cut half the vehicle off.
This is the mistake that costs viewings: the repainted wing, the erased scratch, the smoothed fabric. Name your faults in the instruction, one by one, and require them to stay visible in the result. A blanket ban like change nothing does not work, a precise list does.
Warm light added to the instruction warms the bodywork, and midnight blue becomes royal blue. Always compare the render with the source photo at the same size before publishing, and if the shade has drifted, explicitly ask again for the same colour as the input.
The plate is the one element of the ad that leads straight back to you. The template leaves it blank from the first generation, which is cleaner than a blur added afterwards and stops you forgetting it on the third photo.
An ad with a single perfect picture inspires less trust than one that also shows the faults in close up. Use the workflow for the cover and the overall views, keep your raw photos for the damaged details, and say so in the description.
Cleaning the setting, straightening the framing and redoing the light are photographer's moves, and nobody objects to them. What is risky is erasing a fault on the item being sold, because the buyer finds it on the spot and an ad that hides a known fault turns against the seller. That is exactly the line the template instruction draws.
That is what Flux Kontext is for: it edits the source image instead of generating a new one. Still check the render before publishing, especially the wheels, the headlights and the trim, and rerun while naming whatever moved. One round trip costs a few cents.
Six credits per image, a few cents, and the full workflow produces three of them. The calculator shows the template total before you sign up, and every node displays its cost before you launch.
The template does not blur it, it leaves it blank: the result is cleaner than a smear of pixels and your registration does not circulate. Nothing forces you to hide a plate in an ad, but publishing it lets anyone trace the vehicle, and professional listings have been neutralising it for years.
Yes: replace put the car in front of a grey wall with keep the location, only remove the bin and the parked cars. A real background reassures some buyers, especially for a private sale, and cleaning up is often enough.
On the sharp areas of the source photo, very rarely. On what is blurred, cropped or in shadow it fills in, and that is where an invented wheel or an approximate badge shows up. Shoot the whole item in daylight and the model has nothing to fill in.
Horizontal 4:3 is the historic format of classified ad sites, and it is the one the template cover uses. 1:1 serves the result grids, vertical serves full screen mobile display: the workflow outputs all three from the same photo.
Yes: duplicate the object branch as many times as you need in the same canvas, or simply swap the photo in the Media node and rerun. The instruction stays the same from one item to the next, which is what makes the series consistent. The guide canvas nodes shows how to duplicate a whole chain.
Yes, the canvas runs in the mobile browser: you take the photo, import it into the Media node and generate without touching a computer. The article creating with AI on mobile covers the canvas gestures by finger.
Listing photos: from kerbside snapshot to a picture that sells
Use this workflowFree account, no card required. The workflow opens pre-filled in your canvas.
Flux Kontext
Seedream 5 Pro
Nano Banana