How to Organize 100,000 Photos
By Jay Harb · Published · 11 min read
A library that large breaks the habits that work at 2,000 photos. Here is the order of operations that actually finishes, why the first pass is the slow one, and what the numbers look like on real hardware.
Why 100,000 is a different problem
Organizing 2,000 photos is a weekend of clicking. Organizing 100,000 is not the same job scaled up — it is a different job, because every technique that relies on you looking at each photo stops being available. At one second per photo, with no breaks, 100,000 photos is 28 hours of staring.
So the method has to be: do nothing per-photo that a computer could do for you, and spend your own attention only where a computer genuinely cannot decide.
That reframing is the whole guide. Everything below follows from it.
Step 1: Stop adding to the mess
Before organizing anything, find every place photos are currently arriving and point them at one destination. In practice that is usually:
- Your phone's camera roll
- A partner's or family member's phone
- SD cards from a real camera
- Screenshots (which are photos, and there are always more of them than you think)
- Old drives, old laptops, old backups
You do not have to consolidate them yet. You just have to know the list, because a library that is still growing in five directions can never be finished.
Step 2: Get everything onto one drive
This is the least glamorous step and the one people skip. A photo library spread across four drives cannot be de-duplicated, because you cannot compare what you cannot see at once.
Copy — do not move — everything into one folder tree on one drive. Use plain, boring folder names. Do not try to organize while copying; you will lose track of what has been copied and what has not, and a half-finished copy is worse than no copy.
If the total is bigger than any single drive you own, this is the point at which to buy one. An external hard drive is cheaper than the time you will spend working around not having one, and an external drive works fine as a photo library home if you set it up deliberately.
Step 3: Index before you organize
Now point a photo manager at that folder tree and let it read everything once. This is the slow step, and it is worth understanding why so you can plan around it.
For each file, the indexer has to open it, read its metadata header, work out the capture date and orientation, and produce a thumbnail. On a fast internal SSD this is limited by how fast the CPU can decode. On an external hard drive — where most large libraries actually live — it is limited by how fast the drive's head can move to the next file.
That difference is dramatic. When we measured Pluto Photos' import against a real 21,117-file library on an external USB hard drive, opening each file once and reading a 64 KB header cost 41 ms per file. An earlier version that opened each file four times — once to check it existed, once for GPS, once for dimensions, once for the thumbnail — cost 92 ms per file for exactly the same result. The bytes were not the problem; the seeks were.
Two practical consequences:
Start the first index and walk away. At 40 ms per file, 100,000 photos is about 70 minutes of pure indexing. At 92 ms it is nearly three hours. Either way, it is not a thing to sit and watch.
Do not run other heavy work on the same drive at the same time. A spinning disk gets slower as you add readers, not faster: measured on the same hardware, one reader sustained 33 MB/s and eight concurrent readers managed 21 MB/s between them. If you kick off a backup at the same time as an import, both take longer than running them one after the other.
Step 4: De-duplicate before you touch anything else
A consolidated library is full of duplicates, because you have just merged four backups of overlapping content. Removing them first means everything after this step operates on a smaller, cleaner set.
Filename matching will not do it. The same photo lives on your drive as IMG_4471.JPG, IMG_4471(1).JPG, 20190812_143302.jpg and a re-encoded copy some messaging app made at 60% of the resolution. What you need is perceptual matching — a fingerprint of what the image looks like, so a resized or re-saved copy still matches its original.
Work through the groups from largest to smallest, keep the highest-resolution copy by default, and do not agonize over burst shots. There is a safe workflow for this that avoids the one mistake that actually hurts, which is deleting the original and keeping the compressed copy.
Step 5: Let the machine do the sorting
This is where the "nothing per-photo" rule pays off. Three passes, all automatic:
People. Face detection groups every face in the library into clusters, and you name the clusters — not the photos. Naming 30 people is a coffee's worth of work and it retroactively organizes tens of thousands of photos by who is in them.
Places. Anything shot on a phone carries GPS coordinates, so a map view organizes a decade of travel with no input from you at all.
Subjects. Modern on-device models can match a plain-English description against what is actually in a picture, so "dog on a beach" or "birthday cake" finds photos you never tagged. Here is how that works under the hood — the short version is that it is a similarity search over image embeddings, and it runs on your CPU.
None of these three require you to open a single photo.
Step 6: Add the small amount of structure only you can add
Now, and only now, add human structure — and add much less of it than you think you need. The mistake at this scale is building a 40-tag taxonomy that you abandon in a month.
A tag is worth creating only if you can finish applying it in one sitting and you will actually search for it later. In practice that is usually five to fifteen tags, not forty. Star ratings work better than tags for "this is the good one", because they are ordinal and you do not have to remember what you called anything.
Step 7: Decide what "done" means
A 100,000-photo library is never finished, so define the finish line yourself. A reasonable one:
- Everything is in one place and indexed.
- Duplicates are gone.
- People are named.
- You can find any photo you can describe in under a minute.
That is achievable in a weekend of mostly-waiting. Building a perfect folder hierarchy is not achievable at all, and chasing it is how people end up with a folder called "Unsorted" containing 80,000 files.
What this looks like in Pluto Photos
Every step above maps to something that runs locally: the photo organizer indexes any folder you point it at, the duplicate photo finder uses perceptual hashing, face detection and plain-English search run on your own CPU with no upload, and smart albums keep the sorting going as new photos arrive.
It is a one-time purchase rather than a subscription, and the free tier handles 500 photos so you can test the workflow on a folder before committing a whole library to it. Download Pluto Photos and start with step three on your worst drive.
Part of the Photo Organization Guides, Duplicate Photo Guides, AI Photo Management Guides guides.
Download Pluto Photos for Windows, macOS or Linux, or read more from the blog.