Skip to Main Content

Listing Partner

Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

Listing Partner — Demo video

Listing Partner turns one rental unit's raw bracketed camera exposures into a listing ready to post. I run media for a property manager with about 493 units. Every vacancy needs photos, and today that means a long Lightroom edit for every unit before anyone can see the listing. One command runs the whole pipeline. It merges each bracket into one HDR photo, corrects the lens and straightens verticals. SAM 3.1 finds every window and Muse Spark confirms it, so the light and color can be balanced for the whole room. Muse Spark keeps the best shots, drops duplicates, labels every room, puts them in walkthrough order and checks that every bedroom and bathroom is covered. Then Muse Spark and the open-weight Muse Spark Open each write the listing description and social captions side by side. Every text is checked automatically: every number must match the unit's facts, nothing can be invented, and both models screen for fair housing wording. If they disagree, a person decides. Last is an honest tidy pass. SAM 3.1 finds carry-out clutter, Muse Spark confirms it, and Muse Image removes only that. Every pixel outside the mask is copied from the original and measured, damage is never touched, and if a fill leaves a mark the photo ships untouched with the reason. A renter should see the place as it really is. Everything was built during the hackathon except a read-only AppFolio connector, an existing tool plugged in for public unit facts.

Team