Automatic review · Why it exists

AI Photo QA Review for Real Estate

AI editing is fast — but a listing photo also has to be accurate. A pretty image with a bent door frame, a white hole for a window, or a color cast is a failed delivery. Automatic review exists so those failures get caught before the client sees them.

How the review works

Every processed photo is measured first: brightness statistics catch washed-out or crushed frames, and pure-white pixel share catches blown windows. Where the AI review service is enabled, a vision model additionally reads the image for crooked verticals, blown window views, color cast, and over-smoothing. Findings become plain-language flags on the photo — never a silent pass/fail.

Exposure balance

What it looks for: washed-out frames and crushed shadows, measured as brightness statistics across the image.

Example: “Looks washed out overall — check highlight detail before accepting.”

Window blowout

What it looks for: share of pure-white pixels, the classic sign of a lost window view.

Example: “12% of the image is pure white — window views may be blown out.”

Verticals and geometry

What it looks for: crooked or converging verticals, bent frames and warped lines.

Example: a leaning door frame flagged after perspective correction missed it.

Color consistency

What it looks for: color casts and exposure outliers against the rest of the shoot.

Example: one bedroom running visibly warmer than the other nine photos.

Over-processing

What it looks for: over-smoothing, plastic textures, and crunchy HDR artifacts.

Example: fabric detail smeared into a waxy surface.

A note on property changes: the review measures exposure, windows, lines, color, and processing artifacts in the finished image. It does not compare rooms against the source pixel-by-pixel, so a changed fixture or missing object is still your call — that is what the compare view is for.

Ready to Deliver vs Need Review

Photos with no findings are marked Ready to Deliver and can be accepted in bulk. Photos with findings are marked Need Review: open the photo, read what was found, then accept it, exclude it, or retry. Photos the review couldn't reach are marked for manual inspection — they are never silently passed.

QA-guided retry

Edit → review → finding → retry with QA feedback → re-check. The flag tells you what to change — for example, a blown-window flag points at the window detail setting — and the new result goes through review again. Each photo includes one shared reprocess allowance used by both automatic and manual retries. See the full workflow →

Relevant guide

Real Estate Photo Editing Checklist →

What review does not guarantee

Automatic review flags likely issues. It does not guarantee every issue will be detected, and a suspected property change should always be reviewed before delivery. Final delivery remains under the photographer's control — review is a second pair of eyes, not a replacement for yours.