Real estate photo standards in Australia: what governs quality.

Real estate image quality in Australia sits under several different mechanisms. Listing portals publish technical upload requirements such as dimensions and file formats. Agencies maintain internal style guides. Editing suppliers work to commercial service agreements. Photography awards are judged. VQS, the Visual Quality Score issued by Standard Vision Lab, does something different again. It is an objective, versioned standard that scores the technical execution of a real estate image set from 0 to 100. It is not a regulator, and technical execution is all it measures.

Mechanisms that sit around a listing shoot

Several mechanisms sit around a real estate shoot, and each answers a different question. Being precise about which question each one answers shows what is left over.

Portal upload requirements

Listing portals publish technical upload requirements, covering matters such as image dimensions and accepted file formats. They describe how a file is packaged for a listing page.

Packaging and capture are different things. A file can satisfy an upload requirement and still be soft, blocked up in the shadows, tinted green by light off a lawn, off vertical, or carrying a window clipped to flat white. The file uploads. The photograph is still wrong. A specification for the container says nothing about the state of the image inside it.

Agency style guides

Agencies maintain internal style guides, and a good one is a considered document. Its limits are structural.

The first limit is language. Where a guide describes a look in words, such as bright but not blown, or warm but not yellow, the words carry no threshold. There is no measured value on either side of the line, so a difference of view has nothing to resolve against.

The second limit is reach. A style guide is an internal document. Whether it reaches a contract photographer, an editing supplier, or another office under the same signage depends on the agreements in place with each of them, not on the guide. Once work crosses an organisational boundary, the guide relies on someone else choosing to apply it.

Editing supplier service agreements

Editing suppliers work to commercial service agreements, negotiated between the parties and binding on those parties. Acceptability under one is settled between the client and the supplier: private to them, able to differ from one agreement to the next, and not a result that anyone outside the agreement can read, compare or verify.

Judged awards

Photography awards are judged, and recognition has real value. Judging assesses the entry placed in front of the judges, so an award result speaks about that entry. A measurement of a delivered set is a different object: specific files, all of them, whether or not any would ever be entered anywhere.

What a measurement layer adds

Set those side by side and the difference is one of object: the file as a file, relationships between parties, a chosen entry. A measurement layer addresses something narrower and more literal. These exact files, scored against a stated version of a standard, produced this result.

That is the job VQS is built for. It is an objective, versioned standard for the technical execution of real estate images, issued by Standard Vision Lab (SVL), and it runs on a deterministic runner. The same image produces the same score. A score states the version it was produced against and remains valid against that version. The standard is explained in plain English here, and the full specification text is in preparation for publication and available on request.

How the score is built

Every image is scored from 0 to 100. Each check returns a status, and so does each image: PASS is 90 to 100, FLAG is 70 to 89, and FAIL is 69 and below. Checks are grouped into families, among them Exposure, Colour, Geometry and Semantic plausibility. Named checks include exposure_dark, shadows_crushed, white_balance_windows, colour_cast, grass_saturation, geometry_vertical, window_clipped, cables_visible, ghosting_artifacts and ai_texture_blacks. The plain-English guide to the checks goes through them. Artifacts keeps its US spelling here, because that is how the standard names its own family and checks.

Two design decisions are worth understanding before you read a result.

A set is scored by its lowest image

A set takes the score of its lowest image, not the mean. The arithmetic is the point. A mean can sit comfortably high while a single image sits well below it, and the mean is the number that hides that image. The lowest-image rule reports it instead. There is more in why the lowest image sets the score.

Set consistency is also measured in its own right. VQS applies sigma statistics per grouping (global, interiors, exteriors and room class), with per-image z-score outlier attribution that names the frame dragging the set.

Some faults are gates rather than deductions

A gate is not weighed against the rest of the result: where a fault is gated, it fails the set outright whatever the composite says. Focus is currently gated. In one 19-image set, every image passed individually, the mean was 98 and the lowest was 93, and the set still failed, because two frames were soft. See the focus gate for the detail.

What a run involves

A run takes 12 to 20 original JPEGs from a daytime shoot. Twilight and dusk arrive in v2, which is set out in the daytime-first roadmap. Every image is fingerprinted with SHA-256, and duplicates within a set or against earlier runs are removed.

Certification is automatic. A certifiable composite earns its tier badge with the result: no application, no review panel. Platinum is 100, Certified is 90 to 99, Provisional is 70 to 89, there is no mark below 70, and no mark at all for a set that fails a gate.

The badge is a live widget, embedded with one line of HTML. It links to a public verification register showing tier, composite, covered files and dates, and it opens the full annotated report. It expires 12 months after the last certified run, and expired, suspended and revoked marks visibly say so.

Certifying a pipeline or a single file

Pipeline certification certifies a production workflow rather than one shoot. SVL chooses the input images, the applicant cannot preview, filter or cherry-pick them, and the images must cover a spread of room types and shooting conditions. If the pipeline materially changes, for example by swapping or retraining a model, the certification no longer covers output made after that change until it is re-tested.

Single-image certification sits at the other end of the scale, binding to one exact file by its SHA-256 fingerprint. How certification works covers both.

What VQS is not

VQS is not a regulator. It measures technical execution and nothing else. Composition, styling, aesthetic preference, storytelling, video, floor plans, the accuracy of listing text and legal compliance are all out of scope. Whether a listing or a campaign complies with the law is a question for lawyers and the relevant authorities.

VQS applies to imagery however it was produced, and a conforming runner must not use any declaration of production method as a scoring input. Only SVL issues certification, scored by its canonical runner. Scores from other conforming runners are advisory. There is more in AI edited real estate photos.

What it costs

A subscription is US$9.99 per month. It includes one run each month, keeps the mark live while you are subscribed, and requires a recertification run at least every 12 months. A single run is US$129, with a badge valid for 12 months. Prices are quoted in US dollars, so the Australian amount depends on the rate on the day you pay.

To see how the standard behaves on your own work, try the demo.