Move-in inspection, move-out inspection. Two photo sets, taken months apart, each enriched with KontexVoX — you dictated what you saw while photographing every room. Now you want a report that compares both and lists the differences: what was damaged at move-in and still is, what appeared during the lease, what was repaired.

Doing that comparison by hand, photo by photo, room by room, is tedious and error-prone. But if every photo carries a description, an AI agent can cross-reference the two sets and produce the report automatically.

Prerequisites

Before you start, you need:

1. Two photo sets enriched with KontexVoX: move-in and move-out. The descriptions live in each file's EXIF Description field. See What is an EXIF file?.
2. An AI agent with file-system access: Claude Code, Codex CLI, or Gemini. Not a web chatbot.
3. The photos exported to your computer via AirDrop or the Files app (the channels that preserve EXIF — see How to share without losing the EXIF).
A web chatbot doesn't compare two folders of photos. It processes images one by one, with no access to your files. To cross-reference move-in and move-out, you need an agent that reads both sets locally.

What you dictated: an example

Here's what the EXIF of the same rooms contains, at move-in and move-out:

MOVEIN/kitchen_01.jpg → "Kitchen, right wall. Brown stain under the sink, 20 cm, old. Cracked tile above the counter." MOVEOUT/kitchen_01.jpg → "Kitchen, right wall. Brown stain under the sink still there, more visible. Cracked tile now broken clean, one chip missing. New scratch on the counter, near the hob." MOVEIN/living_03.jpg → "Living room, ceiling. No visible defect. Clean paint." MOVEOUT/living_03.jpg → "Living room, ceiling. Water infiltration mark in the left corner, 30 cm. Appeared recently."

These descriptions, you dictated on site, while photographing. KontexVoX transcribed and engraved them into the file. The AI agent will cross-reference move-in and move-out room by room, and identify exactly what changed — without you having to re-compare everything visually.

The prompt

There are two folders: "movein" and "moveout". Read all the JPEG images in both. For each image, extract the EXIF Description field and the date. Compare the descriptions between move-in and move-out for each room (inferred from the descriptions). Generate a Markdown report listing: defects present at move-in and still present at move-out, defects that appeared between move-in and move-out, defects repaired between move-in and move-out. For each difference, quote the exact descriptions from both photos.

What the AI produces with this prompt

A Markdown report in three sections:

1. Persistent defects — present at move-in, still there at move-out.
2. New defects — appeared between move-in and move-out.
3. Repaired defects — present at move-in, gone at move-out.

Each point quotes the exact descriptions you dictated, with the photo file name as reference. It's a documented, traceable difference report that both landlord and tenant can consult.

The report is in Markdown, readable as plain text and exportable to PDF. You can attach it to the move-out record, send it to the landlord or tenant, or keep it for yourself.

Recommended tool: Gemini

Gemini is good at image analysis and visual comparison. For this case, where the AI must compare two states, Gemini can also use its vision capability to compare pixels beyond the EXIF descriptions. That's a double layer of verification: the text you dictated + the AI's visual analysis.

Frequently asked questions

Is the report legally binding?

The report is a working document. It documents the differences in a traceable, quoted way, but it has no legal force in itself. It's support material for your case — not a substitute for a formal procedure. In a dispute, the original photos with their EXIF metadata (date, location, description) remain the evidence.

Does it work if the move-in and move-out photos aren't shot from the same angle?

It works less well. The prompt asks the AI to compare "for each room (inferred from the descriptions)" — the text makes the link, not the photo's position. If your descriptions mention the room ("Kitchen, right wall"), the AI can pair the photos. If they don't, the AI will try to guess, with possible errors. Hence the importance of always starting a description with the room.

How many photos for a complete inspection?

Generally 30 to 50 photos for a standard apartment: every room, several angles, defects in close-up. With KontexVoX, you describe while walking, photo by photo — 50 enriched photos take 15 minutes. Without enrichment, finding and comparing every defect 12 months later is an uphill battle.

Can I do the same thing without KontexVoX?

Yes — you can type the descriptions by hand into each photo. On a 50-photo inspection, that's already laborious. But the real test is move-out: 12 months later, redoing the same exercise on the same rooms, typing by hand, forgetting nothing. That's where voice changes everything — you describe what you see at the moment you see it, without coming back to it.

Going further

Generating a signable inspection report PDF Generate a structured appraisal report Check that your photos are properly enriched Capture Mode or Classic Mode: which one in the field? Optimize an enriched-photo session Inspections page