Shooting windows that pass: what the standard measures.

Windows put two lighting conditions into one real estate interior: the room, lit by whatever is inside it, and the view beyond the glass, lit by the sky. VQS, the Visual Quality Score published by Standard Vision Lab, is an objective versioned image-quality standard for real estate photography, and it measures the window directly through checks named window_clipped, window_dark, window_artifacts and white_balance_windows. Every check is scored from 0 to 100 on every image, with PASS at 90 to 100, FLAG at 70 to 89 and FAIL at 69 and below. VQS measures technical execution and applies to imagery however it was produced.

Windows are two photographs in one frame

An interior with a window holds two scenes at once. There is the room, lit by whatever is inside it. There is the world outside, lit by the sky. The brightness gap between them is wider than a single exposure covers, and the eye adapts as it moves around a room in a way a photograph cannot.

Expose for the room and the view turns to paper. Expose for the view and the room falls away. Every window technique is an attempt to hold both, and every technique leaves evidence when it is pushed too far.

VQS measures the result, not the method. A conforming runner must not use any declaration of production method as a scoring input. That position is set out in AI edited real estate photos.

What the standard measures at the glass

These check names sit on the window itself: window_clipped, window_dark, window_artifacts and white_balance_windows. Check and family names are quoted exactly as the standard writes them, which is why Artifacts carries an i inside a name while ordinary prose here uses artefacts. The names you can cite are listed in the checks.

window_clipped

The view is gone. Values in the window area have run to the top of the scale and taken the detail with them: no sky gradient, no foliage, no line where the neighbouring roof meets the air. The report names what it measured, so the distinction between a small bright detail and a window with no view left in it is visible in the result rather than argued about.

window_dark

The opposite fault. The window reads as a shape rather than a scene, dark enough that the view outside is not legible. The standard measures this under the name window_dark.

window_artifacts

Halos, edge fringing, a bright or dark rim tracing the frame, a glow around the mullions: the marks left by assembling the window from more than one exposure. The standard measures them under the name window_artifacts.

white_balance_windows

Daylight through glass is not the colour of the light inside the room, and the glass itself tints what passes through it. Neutralise the room and the view can drift cool or green. Neutralise the view and the room turns heavy and warm. The window has a check name of its own here, alongside white_balance and white_balance_tint, so treat the two light sources as a relationship rather than one number: warmer inside, cooler outside, with neither half turning a colour the building does not contain.

The checks a window can drag with it

A window problem does not always land on a check named for the window.

• ghosting_artifacts: doubled edges from merge errors. Leaves in the garden, a ceiling fan, a curtain in a draught, traffic beyond the glass. Anything that moved between frames can leave a soft second copy of itself.
• exposure_bright: the frame lifted overall to keep the interior open after the window was pulled down. The room flattens and the whole image sits high.
• shadows_crushed: the mirror of that move. Blend for the view, and the corners, the underside of the joinery and the far end of the room go to black with nothing in them.
• noise: the bill for lifting a room that was left dark to protect the glass.

Every check is scored per image on the same scale, 0 to 100, with PASS at 90 to 100, FLAG at 70 to 89 and FAIL at 69 and below. A window problem can surface under more than one check name at once.

Bracketing discipline

• Cover the range with no gap in it. The darkest frame should hold the view with a little room to spare. The brightest should open the deepest shadow in the room. Work out your own interval on a room you know, then keep it.
• Change one thing between frames. Shutter only. Moving aperture shifts depth of field and moving focus shifts the edges.
• Remove the movement you control. Close the doors, still the curtains, stop the fan, wait out the wind in the garden.
• Keep the camera locked between frames. Small shifts become edge artefacts at exactly the boundary you were trying to protect.
• Shoot every room the same way. A set built with one method is easier to finish the same way than one assembled room by room.

The set is scored by its lowest image

A set is scored by its lowest image, not the mean, which is covered in set consistency and the lowest image. That changes where your effort belongs. A rescued hero shot in the lounge does not lift a blown window in the second bedroom.

Consistency is measured in its own right as well, with sigma statistics for the set as a whole and for groupings such as interiors, exteriors and room class, plus per-image z-score outlier attribution that names the frame pulling away from the rest. Some faults are gates rather than deductions, and a gate fails the set outright whatever the composite. Focus is currently gated. In one example, a 19-image set in which every image passed individually, with a mean of 98 and a lowest image of 93, failed because two frames were soft.

Where to stop

The instinct on a hard window is to keep working. Leave the light falling off around the aperture the way it actually falls off. Do not paint contrast into the frame edge to make the window feel present. A small clipped specular is a smaller thing than a halo around the whole opening. If a blend needs heavy masking to hold together, the source frames were wrong, and the fix belongs on the shoot rather than in the edit.

Self-check before you deliver

1. View every interior at 100 per cent along the window edge. Halos can hide at fit to screen.
2. Ask of each window: can I name what is outside?
3. Put all the windows in the set side by side and look for the one that is cooler, brighter or flatter than the others.
4. Hunt for doubled edges in anything that could have moved.
5. Check corners and ceilings for black with no information in it.
6. Run the set and read the annotated report. You can see how one reads on the report page.

What the standard leaves to you

A good window treatment is one a buyer would recognise when they stand in the room: the same view, the same relationship between inside and outside. That last judgement is yours. VQS measures technical execution only and stays out of composition, styling, aesthetic preference and storytelling, and it applies to your imagery however it was produced. The runner is deterministic, so the same image returns the same score, and a score states its version and stays valid against it.