You photographed your garage, workshop or stock with KontexVoX in Capture Mode. Photo by photo, you described what you saw: "Crate J02 — Makita drill driver, 2 batteries, charger, case. Bosch underneath." 90 photos, 90 descriptions, engraved in the EXIF. Now you want to turn that into a usable Google Sheets table.

A table is the shape data takes when you want to sort, filter and search it. Your EXIF descriptions are free text. The move to a table — columns, rows, categories — is what the AI agent does for you.

Prerequisites

Before you start, you need:

1. Photos enriched with KontexVoX. 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).
90 photos means 90 descriptions to structure. A web chatbot processes them one by one, with no access to the batch. To produce a complete CSV, you need an agent that reads the whole folder at once.

What you dictated: an example

Here's what the EXIF of four photos from a garage inventory contains:

IMG_0045.jpg → "Crate J02 — Makita drill driver, 2 batteries, charger, case. Top shelf." IMG_0046.jpg → "Crate J08 — hand tools: screwdrivers, adjustable wrenches, wire cutters, tape measure. Bottom shelf." IMG_0047.jpg → "Crate J14 — maintenance products: brake cleaner, WD-40, microfiber cloths. Floor, right of the workbench." IMG_0048.jpg → "Crate #1393 — files 1941-45, sealed metal box, signed envelope. Up top, back shelf."

These descriptions, you dictated while walking through the garage, photo by photo. KontexVoX transcribed and engraved them. The AI agent will extract the information, infer categories (power tools, hand tools, products, archives), and produce a CSV with one row per photo.

The prompt

Read all the JPEG images in this folder. For each image, extract the EXIF Description field, the capture date, and the file name. Generate a CSV file with the following columns: Number, File name, Date, Description, Category (inferred from the description). Sort by file number. The CSV must be directly importable into Google Sheets.

What you get

A CSV file with 90 rows, one per photo. The columns: Number, File name, Date, Description (your dictated text), Category (inferred by the AI: tools, products, archives, etc.).

Import the CSV into Google Sheets: File → Import → Upload. You get a sortable, filterable table. Type "drill" in the search, you find the row. Sort by category, all the tools group together. The table is clean, structured, and searchable.

Adding an "AI comment" column is possible: ask the agent to decode voice-dictation errors (for instance "cross-head goat" → "cross-head screw", "mason's leap" → "mason's bucket"). This column is useful when you enrich in Capture Mode with background noise.

Recommended tool: Codex CLI

Codex CLI (OpenAI) excels at tabular data processing and CSV export. For this data-structuring task, Codex is more direct than Claude Code.

Frequently asked questions

Can I add custom columns?

Yes. Add to the prompt: "Add a Location column (inferred from the description: garage, basement, workshop) and a Value column (left empty)." The AI adjusts the CSV accordingly.

Does it work with more than 90 photos?

Yes, there's no limit. The initial test covered 92 photos, but the prompt works with 200 or 500. The CSV just gets longer. The AI's generation time increases, but the result is the same.

Is the inferred category reliable?

On clear descriptions ("Makita drill", "maintenance products"), yes. On ambiguous ones ("box with stuff"), the AI marks "to verify". You validate in Sheets after import. The category is an aid, not a truth.

Can I do the same thing without KontexVoX?

Yes — you can type 90 rows by hand into Sheets. That's an evening of data entry. Voice changes the method: you describe while walking, in 15 minutes, and the table builds itself afterwards. That's why I created KontexVoX: so the inventory happens at the moment you're looking, not at night in front of a spreadsheet.

Going further

Generate an optimized storage plan for your garage Create a compliant accounting inventory file Create a searchable box index Capture Mode or Classic Mode: which one in the field? Inventory page