Why not just use AI-generated mockups?
Image models can put a screen on a desk in seconds. They still cannot give you a real device, real glass, and a result you can trust in a listing.
It is a fair question. Prompt a model for "laptop on a wooden desk, morning light" and you get something in seconds, for nothing. So why photograph devices, build masks by hand, and write a tool to composite them?
The device is not real
A generated laptop is an average of every laptop the model has seen. The bezel is a little too thick, the keyboard has one key too many, the logo is almost right. People who own the device notice in the first second, and those are exactly the people a mockup is for. A reMarkable owner looking at a planner listing knows what the tablet's edge looks like. A photograph gets it right because it is the object.
The screen is the hard part
A mockup exists to show your design on a screen. Models are worst at precisely that: the screen contents warp, letters smear, and the perspective of the design rarely matches the perspective of the device. You can try to paste your screenshot on afterwards, but then you are doing the compositing work anyway, without a mask, without the reflection, and without knowing the corner points.
Our scenes are the other way round. The photograph is fixed and measured. The screen is a quadrilateral we know to the pixel, the mask follows the real edge of the glass, and the reflection is the light that was actually on it. Your screenshot is warped into that geometry and lit by that reflection. The result is your design in a real photo, not a photo-like guess.
You can use it
A listing image, an ad, or a client presentation needs an image you are allowed to use. With a generated image the answer depends on the model, its terms, and where its training data came from, and that answer changes every few months. With our scenes it is a sentence: renders from the tool are yours to use commercially, including in marketplace listings, ads and client work. The licence says so in plain language.
Consistency
Ten generated images are ten different desks, ten different lights. Ten renders from one photographed scene are the same desk and the same light with ten designs. For a shop, a product line, or a portfolio, that consistency is the point.
Where generated images are fine
Mood boards, a quick internal sketch, a rough idea of layout. We use them for that too. The moment the image has to represent a real product on a real device to a real buyer, the shortcut costs more than it saves.
That is the whole argument. It is not that models are bad. It is that a mockup's job is to be true, and a photograph is true by construction.
Try the mockup tool, free, in your browser.