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.”
Automatic review · Why it exists
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.
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.
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.”
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.”
What it looks for: crooked or converging verticals, bent frames and warped lines.
Example: a leaning door frame flagged after perspective correction missed it.
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.
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.
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.
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 →
Real Estate Photo Editing Checklist →
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.