Skip to main content
Visual AI damage detection

See the damage. Know the cost.

Snap a wall, a carpet, a fixture. The AI flags the damage type, scores the severity, and returns a localized repair estimate in seconds. From the move-out walkthrough or a tenant's maintenance photo, every shot becomes a typed, priced, timestamped record you can defend.

98.4%Detection accuracy
2.6sAvg analysis time
42Damage classes
$380Avg dispute saved
The assessment card

One photo becomes a typed, scored card.

The moment you upload a picture, the vision model turns it into an assessment card. It names the damage type, scores the severity, and puts a localized cost range on it, all from the pixels alone, before you have written a single note.

The damage named across 42 classes, not a one-line guess
Severity scored with a confidence reading on every call
A localized cost range, ready in seconds
The deduction line

Localized materials and labor, itemized.

The assessment does not stop at a round number. Materials and labor are priced for your ZIP, backed by a continuously refreshed corpus of contractor rates and SKU pricing, then totaled into a suggested deduction line you can drop straight onto the security deposit refund.

Materials and labor priced for your ZIP, not a flat guess
A suggested deduction you can itemize, not round off
Posts to the deposit refund as a defensible line item
The court packet

Move-in versus move-out, wear filtered out.

Move-in and move-out photos sit side by side. The AI compares them against the baseline, filters out normal wear and tear, and assembles a court-ready packet (timestamped photos, confidence scores, localized estimates, and a state-specific deduction letter) you can download as a PDF. When a tenant disputes, you walk in with receipts.

Move-in baseline compared to move-out, side by side
Normal wear and tear filtered out of the deduction
A court packet you can download as a PDF
See Online Documents
Snap, scan, document

One loop. The AI does the first two steps.

You snap a photo, the AI scans it, and the documentation writes itself. A phone camera is fine, no fancy lighting required. The vision model handles the typing and the pricing, so the only step left to you is the one that takes a moment.

A phone camera is enough, no special setup
The scan runs in seconds, type and severity from pixels
The deduction letter is generated for you
FAQ

Damage detection questions

How accurate is the damage detection?
The model runs at 98.4% top-1 accuracy, benchmarked against a 12,000-image labeled test set every quarter. Underperforming classes get retrained, so accuracy keeps climbing. Wall scuffs run a little lower because they overlap heavily with normal wear and tear.
Where do the repair estimates come from?
Materials and labor are priced for your ZIP, backed by a continuously refreshed corpus of contractor rate data and SKU pricing. Each detection returns a material average and a labor estimate so the suggested deduction is itemized, not a round number.
Will this hold up in a deposit dispute?
Yes. Each photo is EXIF-timestamped and tied to the unit. The AI compares move-out against the move-in baseline, filters out normal wear, and assembles a court packet (timestamped photos, confidence scores, localized estimates, and a state-specific deduction letter) you can download as a PDF.

Stop arguing about damage. Start documenting it.

Photos in. Estimates out. Audit trail attached. The next move-out walkthrough writes itself, side by side against the move-in baseline.

Free to start. No card required. Upgrade anytime.