Everything you need, in plain English. No tech-speak.
Face Tagger looks at your photos and figures out who's in them. Once it's done, every person is added as a Lightroom keyword — so you can search your catalog by name the same way you'd search for "beach" or "wedding".
You show it a few photos of each person first ("this is Alice, this is Bob"). After that, it can tag the rest of your catalog on its own.
face-tagger.lrplugin and face-tagger-server — need to sit in the same parent folder. This is critical. The plugin looks for the server next to itself. If you separate them, nothing will work.The very first time you use any Face Tagger feature, the plugin has to download its AI files to your Mac. This is a one-time thing. Two things to know:
After that one-time setup, Face Tagger runs entirely on your computer. No internet needed for any feature. Your photos and face data never leave your Mac.
face-tagger-server folder in Finder, right-click the face-tagger-server file inside, and choose Open. Click Open again when macOS asks. You only do this once. See the Troubleshooting section if you hit a persistent block.
If you only read one section of this guide, read this one. Follow it end-to-end and you'll have your first batch of tagged photos done.
Start with the people who show up over and over in your catalog. Parents, partner, kids, closest friends. You'll add more people later — don't try to cover everyone at once.
For each of those 2–3 people:
Alice) and click Next.Nothing visible happens in Lightroom yet — that's normal. Training doesn't tag anything. It just teaches the plugin who these people are. Tagging happens in the next step.
Start small. One trip. One event. A month of photos. Somewhere between 50 and 500 photos is perfect for a first run.
During the scan, the plugin will sometimes stop and show you a grid of face crops with the question "Is this someone you know?" These are groups of similar-looking faces the plugin found but doesn't have a name for yet.
Your options for each cluster:
When the scan finishes, a summary dialog shows you how many photos got tagged. Open Lightroom's Collections panel (left sidebar of the Library module). You'll see a new collection set called Face Tagger Results. Inside, there's a collection for each person you trained, showing all the photos they appear in.
You'll also see the keyword People > [Name] applied to every tagged photo, under Lightroom's Keyword List. From now on you can search by name like any other keyword.
That's it. Your first batch is tagged. To tag more photos, select them and repeat Step 4.
Face Tagger is a loop, not a one-time event. Each pass makes the plugin smarter. The loop has four steps:
Train → Scan → Review → Sync → Repeat
Training is how you tell Face Tagger "this face is Alice". You pick some photos of Alice, give her a name, and the plugin remembers what she looks like.
When to train:
Important: training by itself does nothing visible in Lightroom. No keywords appear, no collections are created. It's preparation. Tagging happens during the Scan step.
Scanning is the step where the plugin looks at photos and applies keywords. Select the photos you want, run a scan, and for every face that matches someone you've trained, a keyword is added under People > [Name].
When to scan: after you've trained at least one person, and any time you want to tag a new batch of photos.
What you'll see in Lightroom after a scan:
After a scan, there will usually be faces the plugin found but couldn't match to anyone trained. Review is a second chance to tag those.
If Review finds unrecognized faces, it opens automatically at the end of a scan (and also at the end of every Deep Scan). You'll see the same kind of cluster dialog as during the scan — grids of similar-looking faces with the question "Is this someone you know?"
You can skip Review entirely. Click Cancel on the dialog. Your scan results are already saved — nothing you've already tagged gets lost. Review is just optional extra tagging.
When to run Review:
When to skip Review:
You can also run Review manually later — see the Review Unrecognized Faces menu item below.
The plugin isn't perfect. It will occasionally tag the wrong person, or miss someone it should have caught. When you notice these, you'll fix them manually in Lightroom (add the right keyword, remove the wrong one). Sync Corrections is how you tell the plugin about those fixes so it gets better next time.
When to run Sync Corrections: after you've gone through a batch of scan results and manually corrected any mistakes. A rhythm that works well:
The plugin counts what you changed and adjusts how strict it is for each person. If you rejected three "Alice" tags, it'll be tighter about matching Alice. If you added three "Alice" tags it missed, it'll be looser. The next scan benefits from what you taught it.
Do you have to run Sync Corrections? No. It's optional. But it's the main way accuracy climbs from "good" to "great" over the first few scan cycles.
Train a new person. Scan a new batch. Review. Sync. Train another. Scan the rest. Review again.
After 2 or 3 correction cycles, accuracy for the people you've trained typically lands in the 90–95% range — and keeps getting sharper with every round.
All Face Tagger features live under Library > Plug-in Extras in Lightroom. Here's each one, in menu order, with what it does and when to use it.
| Your situation | Use this |
|---|---|
| First time, want everything tagged automatically | Deep Scan |
| People are trained, just tag more photos | Scan Selected |
| A folder full of one untrained person | Suggest People |
| Want control over strictness & who's included | Scan Selected |
What it does: teaches the plugin what a person looks like, using photos you've picked as reference.
When to use it: at the start (for the people you care about most), and any time you want to add a new person or improve recognition of someone already trained.
What happens: you select photos, run the menu, type a name, and for every photo with a single clear face the plugin quietly adds that face to the person's training set. If a photo has multiple faces, a dialog appears for it asking you to name each face (or skip the ones you don't want). Training by itself doesn't apply keywords or create collections — that's what Scan is for.
Limits: you can train on up to 20 photos at a time. If you need more, just run Train again — it adds to the existing person.
What it does: runs the full Suggest → Scan → Review flow automatically, in one click.
When to use it: first time scanning a new batch, or whenever you want the plugin to do "everything" without you picking through individual steps. Deep Scan is the recommended option for most users most of the time.
What happens: the plugin first looks for clusters of similar faces (and asks you to name them); then scans against everyone you've trained; then runs Review on anything left unrecognized. You'll see dialogs at each stage where it asks for names — answer what you can, skip the rest. When it's done, you get a summary and collections appear under Face Tagger Results.
Limits: needs at least 5 photos selected, max 2000. If you've got more, split into batches.
What it does: learns from the keyword changes you've made after a scan and adjusts how strict the plugin is for each person.
When to use it: after you've spot-checked a batch of scan results and manually fixed any wrong/missing tags. Run this on the same photos you corrected.
What happens: the plugin compares the current keywords on your photos with what it originally suggested. It counts three things: tags you confirmed (kept), tags you rejected (removed), and tags you added (missed by the scan). A summary dialog appears showing the count per person. If you click Submit, the plugin saves the feedback and auto-adjusts the match strictness for affected people. The next scan uses the tuned values.
Good to know: if you sync the same batch twice, the second sync does nothing — the feedback is only counted once.
What it does: matches the photos you've selected against the people you've trained, and tags whoever it finds.
When to use it: when you've already trained the people you care about and just want more photos tagged, without the extra Suggest/Review steps that Deep Scan adds. Also use it when you want control — adjust the match strictness, exclude certain people from this scan, etc.
What happens: a dialog appears with a few controls:
Click Scan. Collections appear under Face Tagger Results when it's done. If there are unrecognized faces, Review opens automatically (you can cancel out of it).
Good to know: scanning a photo that's already been scanned will skip it. To re-scan after improving training, use Clear All Scan History first.
What it does: finds groups of similar-looking unknown faces across a set of photos and asks you to name them in bulk.
When to use it: when you've got a folder full of someone (or a handful of people) you haven't trained yet. One dialog at a time, you see a grid of faces from the same person and name them all at once.
What happens: the plugin scans your selection for faces, groups similar ones together, and then shows you each cluster with a "Is this someone you know?" dialog. Type a name, or pick one from the dropdown if it's someone you've already trained. After naming, the cluster becomes training data for that person and those photos get the keyword applied. Clusters you Skip are left untagged.
Suggest People is the fastest way to tag a new person you don't have training for yet.
What it does: opens the cluster-naming dialog for any faces left unrecognized after your most recent scan.
When to use it: Review usually runs automatically at the end of a scan. You'd use this menu item manually if you cancelled out of it earlier and want another chance, or if you trained new people since the last scan and want to pass over the unrecognized faces again.
What happens: same as Suggest People — cluster dialogs appear one by one with face grids asking "Is this someone you know?" Name what you want, skip the rest. Tagged faces are added to Lightroom keywords and collections immediately.
Remember: Review is recommended but not required. Clicking Cancel loses nothing — your earlier scan results are already saved.
What it does: looks at a single photo and tells you who the plugin thinks is in it.
When to use it: quick spot checks. "Is this really Alice, or did the plugin get it wrong?" Or: "I don't remember this person — who does the plugin think it is?"
What happens: select exactly one photo, run the menu, and a summary dialog appears listing each face in the photo with the plugin's best guess — e.g., "Face 1: This looks like Alice (strong match). Face 2: This might be Bob (possible match)." No keywords are applied, no collections created. It's purely informational.
Limits: works on exactly one photo at a time, and the original file must be available on disk (not just a smart preview).
What it does: opens a dispatcher dialog with all the "housekeeping" tools for your trained people and scan history. Clicking a button inside opens the matching sub-dialog.
When to use it: any time you need to rename, delete, adjust, or reset something about your training data or scan history.
The buttons, one by one:
Changes a trained person's name. If the new name already exists, you'll be asked to confirm a merge — both sets of training data combine under the new name.
Watch out: this does not update keywords already applied to your photos. If you rename "Ale" to "Alice", photos still have the "Ale" keyword. Lightroom's SDK doesn't let plugins bulk-rename keywords, so you'll need to rename the keyword yourself in Lightroom's Keyword List panel (or leave it as-is and just use the new name going forward).
Removes a person's training data so the plugin no longer knows who they are. A checkbox also lets you delete all people at once (with a confirm prompt).
Watch out: keywords already on photos are not removed — only the plugin's memory of the person. If you want the keywords gone too, remove them in Lightroom.
Lets you adjust how strict the plugin is about matching one specific person, overriding the global default. Pick a person from the dropdown, untick "Use global default", and slide the slider to strict or loose.
When to use it: when one person is consistently getting mis-tagged (make them stricter) or missed (make them looser). Most people should never need this — Sync Corrections tunes these automatically based on your manual fixes.
Removes specific photos from a person's training set without deleting the whole person. Use when you realize a training photo wasn't great (bad lighting, wrong expression, someone you mistagged).
How to use:
Creates (or refreshes) a Face Tagger Training Data collection set showing exactly which photos you used to train each person. Useful for reviewing what the plugin learned from, and essential before using Forget Selected Training Photos.
Creates (or refreshes) an All Scanned Photos collection listing every photo in the catalog that has been scanned at any point. Handy for bulk follow-up actions on scanned photos.
Wipes the plugin's memory of which photos have been scanned, so they can be re-scanned with your current (improved) training data.
When to use it: when you've significantly improved training and want to re-scan old photos to catch matches that didn't hit the first time.
Watch out: this does not remove any keywords or collections. It just resets the "already scanned" flag. Your existing tags stay put; re-scanning will add to them where new matches are found.
What it does: unlocks the plugin using the license key from your purchase email.
When to use it: once, after you install. Until you activate, every Face Tagger feature shows a "please activate" prompt.
What happens: paste the license key into the dialog and click Activate. The plugin confirms with "License activated." and you're done. Don't have a key yet? The dialog has a Purchase a License button that opens the checkout page in your browser.
One license works on two computers at a time (e.g., laptop + desktop).
What it does: removes the license from this computer, freeing up one of your two activation slots.
When to use it: before moving the plugin to a new computer, or if you want to free a slot for a different machine.
What happens: one quiet click, the license is cleared from this Mac. Next time you run a Face Tagger feature, you'll be asked to activate again.
What it does: copies the most recent engine log to your clipboard and opens our feedback form in your browser.
When to use it: anything acts weird. A crash, a scan that stalled, a feature that didn't do what it should. The log contains clues about what went wrong.
What happens: a confirmation dialog shows the path to the log file. Click Open Form. The last 200 lines of the log are silently copied to your clipboard (plain text, no color codes), and the feedback form opens. Paste the log into the form's description field (⌘V) and add a short note about what you were doing when the issue happened. It makes diagnosing ten times faster.
The first time Face Tagger runs, it downloads about 500 MB of AI files to your Mac. This takes 2–5 minutes on a typical connection, and there's no progress bar — Lightroom may look frozen. It's a one-time thing. After that, everything runs locally and fast.
Yes — fully, after the one-time first-run download. All face recognition happens on your computer. No internet is needed after setup. The only exception is license activation, which calls home to verify your key.
Yes. Face Tagger reads .ARW (Sony), .CR2/.CR3 (Canon), .NEF (Nikon), .RAF (Fuji), .DNG, and most other common RAW formats directly. No JPEG export needed.
Not with the downloadable installer — v1 is Mac-only. Windows users can run the plugin from source (we've tested it on Windows 11). Use our contact form and we'll send the setup steps.
Three to eight is plenty. More doesn't hurt, but quality matters far more than quantity. A handful of sharp, well-lit, front-facing photos works better than dozens of blurry or far-away ones.
Yes. If you select a photo with multiple faces and run Train, the plugin will show each face one at a time and ask for a name. Type names under the faces you want to train, and leave the others blank to skip them.
Absolutely. Select new photos, run Train Selected as Person, and use the same name. The new photos are added to the existing training set — the plugin will be better at recognizing that person after each round.
Yes. Open Manage People & Data > View Training Data Collections. It creates a collection set showing each person's training photos.
Use the Forget Selected Training Photos feature. First run View Training Data Collections to see your training photos in Lightroom. Select the bad one(s). Then run Manage People & Data > Forget Selected Training Photos. The specific photos are removed from training; the person and their other training photos stay.
A few possibilities. The face might be at an unusual angle, or lighting/sharpness that's far from the training references. Try adding 1–2 more training photos of that person from similar conditions. Or use Set Person Tolerance to loosen matching for them. Also worth checking: is the original file actually on disk, not just a smart preview?
By design, Face Tagger skips photos it has already scanned. This avoids doing the same work twice. If you've improved training and want those old photos re-scanned, go to Manage People & Data > Clear All Scan History, then scan again.
Strong Match = the plugin is very confident this is the right person. To Review = possible match worth a human glance. Both get the keyword applied, but splitting them into separate collections helps you sanity-check the uncertain ones. You can turn the split off from the Scan dialog if you'd rather have one simple collection per person.
Sometimes a person's face isn't visible in a photo — they're turned away, or the face is too small, or partly hidden. But the plugin can still recognize them from their body (clothing, posture, shape). When that happens, the photo lands in a "Body Match" collection so you know the match was made without seeing a face. Body matching runs automatically on every scan; you don't have to enable it.
The plugin found a person in the photo but couldn't see any face clearly (from behind, very small, blurred, etc.). Those photos land in their own collection so you can review them separately if you want. Body matching may still tag some of them if they match someone trained.
Roughly 1–3 seconds per photo on most Macs. 500 photos is about 10–20 minutes. Speeds depend on your hardware (Apple Silicon is noticeably faster than older Intel Macs) and whether the photos are JPEGs or RAW.
No — closing or quitting Lightroom will abort the scan. But your progress is safe. Any photos that finished scanning are saved. If you resume later, the plugin will skip those and continue where it left off.
No — skipped clusters are remembered for that session. If you change your mind, run Review Unrecognized Faces again manually. Skips reset between sessions if you restart Lightroom, so it's fine to come back to them another day.
No. It's optional. But it's the main way the plugin gets smarter about your photos of your people. Skip it and accuracy stays flat; run it after every scan-and-fix round and accuracy climbs noticeably.
Usually two or three. The first cycle calibrates against real photos; the second cleans up patterns the plugin over- or under-matched; by the third most recurring people are tagged confidently.
This is a Lightroom limitation — the SDK doesn't allow plugins to bulk-rename keywords. Your training data is renamed inside Face Tagger, but photos still carry the old keyword. To fix the keyword side: open Lightroom's Keyword List panel, right-click the old name, and rename it. That updates every photo at once.
The scan itself isn't a single undo-able action, but the Face Tagger Results collections can be deleted safely (this removes the collections, not the keywords). If you want keywords gone too, select the affected photos and remove the keywords in Lightroom's Keyword List panel.
Open Manage People & Data. Use Delete a Person with the "Delete ALL people" checkbox to wipe all training. Then use Clear All Scan History to wipe the scan log. Delete the "Face Tagger Results" collection set in Lightroom. Remove the People keywords from the Keyword List if you want a fully clean slate. You can then train fresh.
Two. One license activates on up to two machines at the same time. If you need a third, deactivate one of the existing ones first — or use the contact form if you've lost access to an old machine and can't deactivate it.
Check your purchase-receipt email from LemonSqueezy — the key is right there. If you can't find it, reply to the receipt for a resend, or use our contact form and we'll look it up.
Use Library > Plug-in Extras > Report a Problem — it auto-copies the recent engine log to your clipboard and opens our feedback form. Just paste the log into the description. For feature requests or general questions, the contact form on the home page is fine.
No. All face recognition happens locally on your Mac. Your photos, face data, and names never leave your machine. The only network call Face Tagger makes is to validate your license key. See the Privacy Policy for the full legal language.
The plugin couldn't reach its AI server. First try: quit Lightroom (⌘Q) and reopen it — the server starts automatically when you run any Face Tagger feature.
If it persists: open Finder and check that the face-tagger-server folder is sitting right next to face-tagger.lrplugin in the same parent folder. If it's not, move it there.
This is Gatekeeper protecting you from unsigned apps. Open the face-tagger-server folder in Finder. Right-click the face-tagger-server file inside and choose Open. Click Open again when macOS asks. You only need to do this once.
That's the one-time AI download. Give it 2–5 minutes on first run. It only happens once. If you're still stuck after 10 minutes, check your internet connection, then restart Lightroom.
Manually fix the wrong keywords in Lightroom (remove the wrong, add the right), then run Sync Corrections and Auto-Tune on those photos. The plugin will tighten matching for the offender next scan. Over 2–3 cycles, false positives drop off.
Split the selection into batches. The limit is there to keep memory use reasonable. Scanning 1,500 photos at a time works fine; just don't throw 20,000 at once.
The photo's source file isn't on disk — maybe the external drive is unplugged, or the file was moved. Face Tagger needs the full original to read faces (smart previews don't have enough detail). Reconnect the drive or locate the original, then try again.
Use Library > Plug-in Extras > Report a Problem — it auto-copies the recent engine log to your clipboard and opens our feedback form. Paste the log into the description and tell us what you were doing. Solo developer, personal responses, usually same-day.