Face recognition for Lightroom Classic: $39 vs $199 vs $270

By David · July 2, 2026 · Updated July 13, 2026 · 12 min read

Full disclosure before anything else: I build Face Tagger, one of the three tools in this comparison. I'm biased, and pretending otherwise would be silly. What I can do is get the facts about all three right, tell you honestly where the other two are better, and let the prices argue for themselves.

What you're actually shopping for

"Face recognition for Lightroom" hides three different problems, and the right purchase depends on which one is yours:

• "Who is in this photo?" You want every shot of your daughter, or of the bride, findable by name. The output you need is keywords in your Lightroom catalog.
• "Tag and cull everything for me." Faces are one part of a bigger wish: autotagging scenes and objects, checking focus, picking the sharpest frame from a burst, maybe tagging video too.
• "My library is scattered across five apps." Lightroom, Apple Photos, Capture One, folders of exports. You want one search box over all of it, and faces are one of the things it should understand.

Each of the three tools was built for exactly one of these. That's why the prices differ by an order of magnitude, and it's why "which is best" is the wrong question. Here's the honest map.

The three tools at a glance

Excire Search 2026: the do-everything plugin ($199)

Excire Search 2026 is the established name here, and it earns the reputation. It shipped in September 2025 at $199 for a full license, runs on both Mac and Windows, and processes everything on your machine.

Face recognition is one feature in a long list. You can search for people by picking an example photo, filter faces by age or gender or how many people are in the frame, and have the results written into your Lightroom catalog as keywords. Around that sit the features that actually justify the price: automatic keywording of your whole library, free-text prompt search, AI-assisted culling with face and eye sharpness detection (green label for sharp faces, red for soft), a duplicate finder, and keywording for video clips.

Two practical notes from the spec sheet that reviews tend to skip. First, it wants resources: 8 GB RAM minimum, 16 GB recommended for big catalogs, and roughly 800 MB of index storage per 100,000 images. Second, upgrades between major versions are paid ($69 from the 2024 version, $99 from 2022), so "one-time" means one-time per generation.

Buy Excire if faces are just one item on your wishlist and you'd use the culling, autotagging, or video search weekly. For a working wedding or event photographer on Windows, it's realistically the only option in this list anyway. If you're weighing it against Adobe's own tools, I've written about what Lightroom Classic's built-in AI culling can and can't do.

Peakto: the library hub that happens to tag faces ($270)

Peakto, from the French company CYME, is a different kind of product. It's not a Lightroom plugin; it's a native Mac app that indexes your Lightroom catalog, Apple Photos library, Capture One sessions, and plain folders, then gives you one search box over all of it.

Its face tagging is genuinely nice: the AI groups matching faces across your whole indexed universe, you name a cluster once, and it tags every match in one shot. Analysis runs entirely on your Mac. It also reads the face tags you already made in Lightroom or Apple Photos, so you don't start from zero.

The catch is directional: Peakto pulls tags in, it doesn't push keywords back into your Lightroom catalog. The names live in Peakto's own database, because the product's whole point is being the layer above your apps. If your goal is Smart Collections in Lightroom Classic keyed to people's names, that's not what this buys you. Pricing reflects the bigger ambition too: the Standard lifetime license is $270 (including one year of major updates), or subscriptions from around $9.99 a month. CYME also sells a separate Peakto Search plugin for Lightroom Classic at $129, but note that one is semantic search only - the face recognition lives in the main app.

Buy Peakto if your photos are genuinely scattered across multiple apps and the "one search over everything" idea makes your shoulders drop with relief. Mac only.

Face Tagger: just the people part, done cheap ($39)

This is mine, so judge accordingly. I built Face Tagger because I wanted one thing Lightroom Classic wouldn't give me: type my kid's name into the Library filter and see every photo of her, including the ones where she's facing away from the camera.

The workflow is deliberately narrow. You pick a handful of reference photos of a person and train the plugin on them - about five minutes. It then scans your catalog locally and tags every other shot of that person as a real Lightroom keyword. Because the model considers more than the face itself, it catches people in profile, from behind, or half out of frame, which is exactly where Lightroom's own People View gives up. There's also a Suggest People mode that clusters unknown faces and proposes groupings, so you can name people you haven't trained.

The reason it holds up on a big catalog is how a person is stored. People View keeps buckets of visually similar face crops; Face Tagger computes a 512-dimensional ArcFace embedding for every face - a numerical fingerprint of identity that survives changes in lighting, angle, and expression - and then averages a person's training photos into a single centroid. Recognizing a new face is one distance comparison per known person, which stays fast and stable whether you have named 3 people or 30. It also reads RAW files directly rather than leaning on Lightroom's preview cache, and it uses a signal Lightroom ignores entirely: photos taken within about five minutes of a confident match are far more likely to contain the same people, so nearby frames get a small nudge toward that person. In sequence shots and event work, that alone catches a lot of faces that would otherwise be missed.

The workflow that scales looks like this:

1. Train first, scan later. Pick 10-20 clear, varied photos of each person. The plugin builds a centroid for each one before it touches the rest of your library.
2. Scan in batches. Run a scan on one shoot or date range at a time. Photos with no person visible are skipped before the face model ever runs, so a landscape-heavy catalog costs you almost nothing.
3. Review unknown faces after each scan. Unrecognized faces come back as tight clusters of 3-15 similar shots, not an infinite unnamed pile.
4. Sync your corrections. Fix a wrong tag in Lightroom, then let the plugin read those confirmations and rejections back in. It tightens recognition for people who get over-matched and loosens it for people who get missed. That correction loop is the part most face tools skip, and it is the part that compounds.

The honest limits: it does nothing except people (no scene tagging, no culling, no video), and v1 ships for Apple Silicon Macs only - no Intel, no Windows yet. First run downloads about 500 MB of AI models, and after that it's fully offline. Everything lands as ordinary keywords, so Smart Collections, filters, and exports just work, and the tags survive independently of the plugin. If you use XMP sidecars, the names even travel with your RAW files.

Buy Face Tagger if the only sentence you care about is "find every photo of this person, as a keyword, in Lightroom Classic" and you're on an M-series Mac. It's a fifth of Excire's price because it does a fraction of Excire's feature list - the one part this post is about.

Which one should you buy?

Here's the decision as I'd give it to a friend over coffee:

And one more option that costs nothing: Lightroom Classic's own People View. Try it before you spend money. If it frustrates you, you'll know precisely which of the three problems above is yours - and it's worth understanding exactly how it frustrates people, because that's the gap all three paid tools are selling into.

Before you spend anything: why People View drifts

You probably noticed it the same way I did. Lightroom's face recognition works well on the first few hundred photos, then starts misfiring. It suggests the wrong name with alarming confidence. The unnamed pile grows faster than you can clear it. Eventually you stop opening People View at all.

It isn't your library organization. Lightroom doesn't build a per-person model the way a modern neural network would - it detects faces, crops them, and groups visually similar crops into clusters, then asks you to name the cluster. Suggestions are "this face resembles the faces already in bucket seven," not a robust identity score. That works fine on a small, consistent set. Add a few thousand photos across different lighting, angles, ages, and image quality and the cluster boundaries start to overlap. Siblings, family resemblance, or two people with similar hair under studio light are exactly where it gives up.

Which leads to the counterintuitive part: the more you use People View, the worse it can get. Every false positive you confirm teaches the cluster that those two faces belong together, and future suggestions inherit the contamination. There's a scale problem too - background face detection runs as a low-priority task, and on libraries over 20,000 photos it can thrash, re-scanning folders it already processed and surfacing photos you tagged years ago as unnamed. And there's the neglect factor: People View shipped in Lightroom Classic 6 back in 2015, and the detection model hasn't had a meaningful public update since. The rest of the field moved on to deep metric learning. Lightroom Classic didn't follow.

Quick fixes worth trying first

If you're not ready to buy anything, a few habits will wring real accuracy out of the free feature:

• Don't let background detection run unattended. Turn off automatic face detection in Catalog Settings and run it by hand on specific folders when you're actually ready to tag.
• Reject bad suggestions aggressively. A face left unnamed costs you nothing. A face confirmed wrong pollutes the cluster permanently.
• Use "Find Faces Again" (right-click a photo in People View) to force a fresh detection pass on images where the first scan missed faces.
• Tag in batches by shoot. Suggestions improve inside a single event, where lighting and framing are consistent.
• Keep the confirmed set clean. Remove faces that are blurry, tiny, or half-obscured. One bad example can skew a whole cluster.

These are workarounds, not fixes. Under roughly 5,000 photos and three or four people, People View is genuinely enough and you should keep your money. Past that - or past a dozen or so named people - the degradation compounds faster than the convenience is worth, and that's the point at which one of the three tools above starts paying for itself.

The part nobody puts in the comparison table

All three tools process faces on your own machine. No uploads, no cloud face database, no terms-of-service change two years from now deciding your family photos are training data. In 2026, when most "AI photo" products are thin wrappers around a cloud API, all three companies independently decided local processing was the only acceptable design for face data. Whatever you buy from this list, you're not trading your archive's privacy for search convenience.

The real difference is philosophy. Excire believes you want a Swiss Army knife inside Lightroom. Peakto believes Lightroom is one drawer in a bigger cabinet. Face Tagger believes you want one sharp blade and $180 left in your pocket. All three are right - for different people.

FAQ

Does Lightroom Classic have built-in face recognition?

Yes. People View has been in Lightroom Classic for years, and it's free. It's also slow on large catalogs, needs a lot of manual confirmation, and has a habit of silently stalling - which is why the plugin market exists. If your catalog is small and you have patience, try it first.

Why does Lightroom's face recognition get less accurate the more I use it?

Because People View matches faces by grouping visually similar crops into clusters rather than building a per-person model. Every wrong suggestion you confirm teaches that cluster the wrong lesson, so the noise accumulates. Large catalogs make it worse - background detection can thrash above roughly 20,000 photos - and the engine itself hasn't had a meaningful update since 2015. Rejecting bad suggestions instead of confirming them, and keeping blurry or half-obscured faces out of the confirmed set, slows the drift down.

Do any of these tools upload my photos to the cloud?

No. All three - Excire Search 2026, Peakto, and Face Tagger - run their face recognition locally on your machine. None of them send your photos or face data to a server. That's rare in 2026 and worth appreciating.

Which face recognition tools work on Windows?

Only Excire Search 2026 (Windows 10 64-bit or Windows 11). Peakto is a native Mac app with no Windows version. Face Tagger currently ships for Apple Silicon Macs (M1 or later) only.

Where do the people tags actually end up?

Excire and Face Tagger both write real Lightroom keywords into your catalog, so the names work in Smart Collections, the Library filter bar, and exports. Peakto keeps tags in its own catalog and imports existing face tags from Lightroom; its purpose is searching across apps, not writing back into Lightroom's keyword list.

Is Face Tagger's $39 price a subscription?

No, it's a one-time purchase with a license key. Excire Search 2026 is also one-time ($199, with paid upgrades between major versions). Peakto offers both subscriptions (from around $9.99 a month) and a lifetime license ($270 for the Standard tier, which includes one year of major updates).