Organization
The problem is not the number, it is the scattering
Forty thousand images are not unmanageable in themselves. What makes a large photo archive hard to catalog is that nobody knows exactly what it holds: external drives piled up over twenty years, memory card dumps, exports made for a publisher, “to sort” folders copied from one computer to the next.
A mid-career artist, a studio that has reorganised itself several times, an estate that inherits a stack of drives with no instructions: the situation is the same. So is the temptation, which is to start describing.
That is the most expensive mistake. You describe files you will later delete, you annotate the same photograph twice under two names, and six months in you discover a whole drive nobody had looked at. Cataloging an archive of this size takes an order: inventory, deduplicate, consolidate, and only then describe, in batches and by priority.
Step 1: inventory before touching anything
The inventory answers one question: what do I have, and where? It is read-only. Nothing is renamed, moved or deleted until the map is complete.
Three criteria for a useful inventory:
- Every medium, including the oldest. They often hold the only copies of the early years.
- Every file that matters: the RAW files, but also working TIFFs, delivered JPEGs, and the XMP files that carry adjustments and keywords.
- Unreadable files recorded, not ignored: a file that is corrupt on one drive may still be intact on another.
Our article Where to start when twenty years of photographs are scattered across drives covers this first mapping in detail.
Step 2: deduplicate by content, not by name
A file name proves nothing. IMG_0042.CR2 exists on every card from every camera; conversely, one image can carry five names after five exports. The only reliable test that two files are identical is a fingerprint of their content.
Then two cases need telling apart:
- the exact duplicate: same bytes, same fingerprint. One copy is enough in the working library; backup copies are a separate matter.
- the visually similar image: a JPEG made from a RAW, a smaller version, a black-and-white conversion. Not a duplicate in the strict sense, and often a piece of information: the version delivered to a magazine tells you something.
The trap is letting software decide on its own. A duplicate decision should stay human, reasoned and reversible. See also What is image deduplication.
Step 3: consolidate into a clean library
Once the decisions are made, gather the files you keep in one place. Copy rather than move, at least the first time. Verify each copy before calling the job done. Keep a record of where every file came from.
This step shapes everything after it. A catalogue relies on file locations: catalog first and reorganise folders later, and part of your description ends up attached to paths that no longer exist.
Step 4: import and describe in batches
With a consolidated library, description can begin. Do not aim for a perfect record for every image. Describe in coherent batches: a series, a commission, a year, an assignment.
- What the batch shares (author, rights, series, technique) is entered once.
- What belongs to each image (title, exact place, people) comes next, where it is useful.
- A controlled vocabulary stops “Marseille”, “marseille” and “Marseille (13)” from becoming three different places.
Start with a batch you know well. It shows you which fields you actually use before you apply them to everything else.
Step 5: set priorities
Not every image deserves the same effort. In an archive of tens of thousands of files, a few hundred are works, a few thousand document their context (contact sheets, variants, installation views), and the rest is working material.
A sensible order:
- images that were exhibited, published or sold;
- series the artist identified as such;
- images that exist as physical prints;
- everything else, over time.
Five hundred reliable records are worth more than a thin layer of approximate keywords over forty thousand files.
Keep a simple log as you go: which drives were inventoried, which batches are described, what remains. A project of this size lasts months, sometimes years, and often passes from one person to another. The log is what lets someone else pick it up.
Step 6: structure into works
The last step separates the image from the work. A file is a digital object; a work is a citable entity with a number that never changes. One work can have several files (scanned negative, retouched master, exhibition view), and a lost work may have none. Giving a catalogue number to every JPEG makes the catalogue unusable. Our article How to build a photographic catalogue raisonné sets out these distinctions.
Pitfalls to avoid
| Pitfall | Consequence | Remedy |
|---|---|---|
| Describing before deduplicating | The same work done twice | Inventory and review first |
| Deleting duplicates straight away | Useful versions lost | Decide file by file, copy before removing |
| Reorganising after cataloging | Orphaned descriptions | Consolidate, then import |
| One work per file | A bloated, confusing catalogue | Create works from a considered selection |
With the Obscura Flow suite
Inventory and deduplicate: Obscura Trace
The Obscura Trace Scan Assistant declares several drives or folders to analyse and only reads them: it writes nothing, moves nothing and deletes nothing there. You choose the file types; XMP is not checked by default, so check it if your RAW files have sidecars. Running an analysis again only rereads new or changed files.
Trace separates exact duplicates (an identical SHA-256 fingerprint) from visual duplicates, which recognise recompression, resizing and black-and-white conversion, but neither cropping nor rotation. In the Review Center, each group shows a recommended master and explains it criterion by criterion; you decide: choose as master, keep, mark as version or mark as duplicate. None of these actions changes a file.
The Transfer Center then consolidates a source into a destination: dry run, plan, confirmation, then a copy verified by fingerprint. The default mode is Copy. A RAW and its XMP travel together; a file marked as duplicate stays in the source. Each transfer writes a manifest that lets you find out, years later, where each file came from. Note that a file the analysis did not inventory is not transferred.
Import, describe, structure: Obscura Catalog
Importing a folder into Obscura Catalog copies nothing: each image is referenced where it lives, with its SHA-256 fingerprint, which recognises a file already present even under another name. An import can be stopped and resumed. Catalog recognises a folder by its path: a folder moved or renamed and imported again becomes a new project. That is a concrete reason to consolidate before importing.
For batches, Catalog offers smart collections and saved searches, tags applied to a selection, and a thesaurus with synonyms. An inventory already kept in a spreadsheet comes in through the CSV import, with a preview before anything is written.
Works are created from a selection of files: one work per file, or a single work for the whole selection. Their numbers are never reassigned. Batch editing changes series, technique or support across several works, with a preview and an undo. To set priorities, each record’s completeness runs from 0 to 100%, the list sorts by completeness, and the Next: missing field link takes you to the criterion worth the most.
Where to start
If your images are spread over several drives, start with Obscura Trace: a read-only inventory commits you to nothing. Once the library is clean, Obscura Catalog takes over to describe it and structure it into works. The suite’s software and the documentation will help you choose.