Interior visualisation from a single photo: how it works and when it fails

Interior visualisation from a photo is now standard equipment in property listings, and one of the most misunderstood services on the market. Some expect an interior design project with a budget and a furniture list; others treat it as manipulation meant to mislead buyers. It sits in between: it re-dresses an existing room so it reads on a photograph.

What interior visualisation is and is not

The confusion starts with the word: "visualisation" describes two entirely different things.

In the architectural sense it is the output of a design: an architect measures the space, plans the layout, selects materials and renders an image from a three-dimensional model. That takes weeks and costs thousands.

Visualisation from a photo starts instead from an existing shot of an existing room. The geometry stays, the windows stay, the doors stay. What changes is the contents: a sofa, a bed, a rug, a few plants. It is dressing, not rebuilding — which is why it takes minutes, and why it shows how a space could be furnished, not how it could be rebuilt.

How it works technically

The model receives two things: a photograph and an instruction — room type and style.

From the photograph it infers the spatial situation: where the floor lies, where the walls begin, how the lines converge and where the light comes from. Get the perspective wrong and the sofa appears to float; get the light wrong and shadows fall the wrong way, which the eye notices instantly.

It then places the furnishings — not from a catalogue of real products, but generated to match the room type and style. Every run is a separate attempt, so two generations from the same photo give two different interiors. That is a property, not a bug, so never work with the first result.

What to expect room by room

Living room. The most reliable room of all: usually the largest, with a continuous floor area. The recurring problem is scale — in a small room the model may propose a sofa set that would not fit.

Kitchen. The hardest room in the flat. Units are fixed construction with precise geometry; the model either preserves them exactly or regenerates them, and then different door fronts appear or a run of units ends in the wrong place.

Bedroom. Works well: a bed is a large, simple object and the room is usually rectangular. Watch out for fitted wardrobes, which the model replaces with free-standing furniture that does not exist.

Bathroom. Difficult: small, full of glossy surfaces, mirrors and sanitary ware. Reflections rarely match what stands in front of the mirror, so a bathroom is better photographed than visualised.

Children's room. Treacherous on age: without further detail the model settles between toddler and teenager. Specify the age.

Home office. An easy room — desk, chair, shelves, lamp — and a good way to give a small spare room an obvious purpose.

Loft and attic rooms. Sloped ceilings are the hardest geometric task there is, because the angle has to be estimated from one frame. The usual result is a wardrobe passing through the slope, or a bed with its headboard in the ceiling.

Cellar, garage, utility spaces. Nothing to offer here. Nobody expects a cellar to look cosy; they want square metres and a dry wall.

How it differs from an interior designer

A designer settles whether something fits in the space. Visualisation settles whether the space looks lived in on a photograph. Only the second question concerns the sale.

You need a designer when work is going ahead: walls coming down, layouts changing, a renovation being costed. That output carries dimensions, materials and prices.

It offers none of those guarantees. For a renovation project it is worth showing both: the real condition and, beside it, the potential. For plots and unfinished houses, visualising a house on the plot does the equivalent job outdoors.

When AI fails

Crooked and sloping walls. Older buildings rarely have right angles. When walls converge, bow or turn into a slope, the geometry is estimated wrong and furniture stands askew or passes through the wall.

Backlight from a window. A photo shot into the window is almost always lost: the window is blown out, the rest dark, the light impossible to infer. The fix is photographic — stand so the window is to your side or behind you.

Clutter in the room. Leftover boxes, ladders and paint tins are handled unpredictably: they either stay in the middle of the new seating group or are smeared into a blur.

Mirrors and large glass surfaces. A mirror should reflect what stands in front of it, including the new furniture. That rarely lands; glazed doors and glossy tiles behave the same way.

Very small spaces. A few square metres leave nowhere to put furniture, but the model tries anyway — you get a bathroom with a tub across the doorway.

Fisheye and ultra-wide lenses. An extremely wide angle curves the lines so the room looks larger than it is. The model does not recognise the distortion and drops straight furniture into it.

All six share a cause: visualisation fails on the source photo, not on the model. The shooting procedure is in the article on furnishing an empty flat before sale.

Labelling the visualisation in a listing

This is not a recommendation and it is not optional. An edited photograph in a property listing must be labelled as edited — portal rules require it, and so does the consumer-law principle that a buyer must not be misled about what is being sold.

Decent labelling means a caption in the image itself, legible on a phone screen, the visualisation sitting in the gallery next to a real shot of the same room rather than replacing it, and unambiguous wording: illustrative visualisation, furniture not included.

The practical risk outweighs the formal one. Formally, an unlabelled edit can be judged a misleading practice; practically, a buyer who discovers at the viewing that the photos showed something else loses trust in the whole offer.

Cost and turnaround

Prices here are orders of magnitude and vary by provider. Visualising one photo costs cents to a few euros and generation takes minutes, so a full set from a flat is done in an afternoon, repeat attempts included. The real bottleneck is the hour you spend photographing, not the computation. Current rates are in the price list.

Turning a visualisation into a video tour

Furnished photos need not end in the listing gallery. The same set can be assembled into a property video tour — stills get motion, transitions appear between rooms, and music, narration and footage of the surroundings are added.

Furnished interiors read as more lived-in in video, because camera movement gives depth. The labelling rule applies there too: the note belongs in the frame.

Summary

Visualisation from a photo dresses an existing room; it does not design one. Living rooms, bedrooms and home offices come out well, kitchens and bathrooms are a struggle, cellars are not worth the effort.

It fails predictably: on crooked walls, in backlight, in clutter, on mirrors, in spaces that are too small and on fisheye distortion. What the results look like in practice is shown under AI visualisation.