Category guide · Updated September 2026

How to find every photo of one person on your computer — offline face recognition apps compared (2026)

The quickest way to find every photo of one person on a Mac or Windows PC, without uploading anything, is FolioSort's Find a Person: add 1–3 reference photos, and it finds every other photo of that person across your folders and drives, sorted by confidence, then copies the ones you keep into a folder. With no photo to start from, People groups the faces automatically and you name them afterwards. Both run on-device — Core ML on the Mac's Neural Engine, local ONNX models on Windows — and names can be written to XMP sidecars for Lightroom, Capture One or digiKam.

It is part of FolioSort Elite, a one-time €39.99, with a free 20-photo taster to check it recognises your family before you pay.

FolioSort is our product. Rival facts were checked on each vendor's own site on 23 September 2026; nobody paid to be listed · No affiliate links

The short answer

  • Our pick: FolioSort — best for finding one person from a reference photo across folders and drives, and turning the result into a real folder, on Mac and Windows, fully offline.
  • Best free, catalog-based: digiKam — open source, local face recognition on Windows, macOS and Linux, if you are happy to keep your library in its database.
  • Photos already in Apple Photos: Apple Photos People — built into every Mac; nothing to install if your photos already live in the Photos library.
  • Self-hosted with phone access: Immich — facial recognition on a server you run, with phone apps, if what you really want is Google Photos without Google.

Also covered: Mylio Photos, Excire Foto, Tonfotos, RetinaTag and PhotoPrism.

The quick list

  1. FolioSort — best for finding one person from 1–3 photos and copying them into a folder · €39.99 one-time (Elite) · free 20-photo taster
  2. digiKam — best free option, catalog-based · free, open source
  3. Apple Photos — best if your photos already live in the Photos library · free, built into macOS
  4. Immich — best self-hosted option with phone access · free, open source, needs a server
  5. Mylio Photos — best if you want your library synced across phone and computer · $240/year subscription
  6. Excire Foto 2027 — best if you want a local AI catalog with prompt search · €229 one-time
  7. Tonfotos — best for family archives and family trees · free version; $59 / $159 one-time
  8. RetinaTag — best no-cost newcomer, if you accept a young project · free
  9. PhotoPrism — best self-hosted alternative to Immich · free community edition; paid memberships

Which offline face recognition app does what?

The columns are the questions people ask when they want every photo of one person out of a big library.

App Runs on Where face work happens Search from a reference photo Groups faces automatically Copy results to a real folder Writes names to XMP Price
FolioSort macOS 14+, Windows 10/11 On your computer Yes, from 1–3 photos Yes (People) Yes Yes, as sidecars €39.99 one-time (Elite); free 20-photo taster
digiKam Windows, macOS, Linux On your computer — Yes (face recognition) — — Free, open source
Apple Photos macOS — — Yes (People album) — (photos live inside the Photos library) — Free, built in
Immich Your own server (Docker) On your server — Yes (clustering) — — Free, open source
Mylio Photos macOS, Windows, iOS, Android On your devices (offline) — Yes (face recognition) — — $240/year subscription
Excire Foto 2027 macOS, Windows On your computer — Yes (face recognition) — — €229 one-time
Tonfotos Windows, macOS, Linux On your computer — Yes (face recognition) — — Free version; $59 / $159 one-time
RetinaTag macOS (Apple Silicon), Windows On your computer — Yes (clustering) — — Free
PhotoPrism Your own server (Docker) On your server — Yes (face recognition) — — Free community edition; paid memberships

“—” = not stated on the vendor's own pages on 23 September 2026. It does not mean the app cannot do it, only that we could not confirm it. Prices as listed that day.

Of the nine, FolioSort is the only app in this list whose vendor documents all three steps of the job — search from a reference photo, copy the matches into a real folder, and write the names to XMP sidecars — on both Mac and Windows.

Why is “find this person” different from “looks similar” search?

A “similar photos” search compares whole pictures. Give it your son on a beach in a red jacket and it returns beaches, red jackets and the same framing — often with nobody you know in them.

Identity matching looks only at the face. FolioSort detects and aligns each face, then turns it into a 512-dimension identity vector with an ArcFace-class model (AuraFace); two faces match when the identity is close, whatever the clothes, place, lighting or crowd. That is how it finds the group shot where someone is half-hidden at the back.

The hard cases are handled honestly: faces too small, turned away or dark to identify are rejected rather than guessed at, and because siblings are where any matcher is weakest, results are confidence-sorted and one correction sticks.

For scene-level search — “kids at the beach”, text inside a photo — see FolioSort's AI Photo Search and the best offline AI photo search apps.

Where do your faces go? The privacy difference

Face recognition computes a set of numbers from each face that identifies that person, much as a fingerprint does. That is biometric data, which European data-protection law (GDPR, Article 9) treats as a special category when used to identify someone. And a family library is full of other people's faces — children, wedding guests, a school team — none of whom agreed to be processed on someone else's server.

So ask where the face work runs. Cloud services such as Google Photos group faces on the photos you have uploaded to their servers. Every app in this comparison except Apple Photos documents that its face work runs on hardware you control: your own computer (FolioSort, digiKam, Excire Foto, Tonfotos, RetinaTag), your own devices offline (Mylio Photos), or a server you run (Immich, PhotoPrism). For Apple Photos we did not re-verify the details for this page, so we make no claim either way.

In FolioSort: no account, it works offline, the face index lives on your own disk, and Clear index deletes the face data without touching a photo.

The nine apps, and who each one is for

1. FolioSort — best for finding one person and getting a folder of them

OURS · MACOS 14+ (APPLE SILICON + INTEL), WINDOWS 10 1809+ / 11 (X64) · ELITE €39.99 ONE-TIME, 3 DEVICES ACROSS MAC + WINDOWS · FREE 20-PHOTO TASTER

A native Mac and Windows photo and video organizer that works on your real folders — no catalog, no import. Find a Person finds everyone matching 1–3 reference photos across the folders and drives you indexed; People groups faces with no example photo and can split a library into one folder per person. The models (YuNet finds faces, AuraFace compares them) are bundled in the app.

Reasons to choose it:

  • Identity matches from a photo of the person, not look-alike scenes.
  • Copy to Folder gives you a real folder and leaves every original in place.
  • Smart Folders can narrow a saved search to one person, so a folder of your daughter keeps filling itself.
  • Names go into XMP sidecars that Lightroom, Capture One and digiKam read; RAW files are never modified.
  • One-time price, used in 80+ countries, 12M+ files processed, a 14-day full refund with no conditions, and the independent developer answers support e-mail.

Reasons to skip it: desktop only — no phone app or web gallery. Not an editor or a keyword-tagging catalog. The free plan is a taster (20 photos copied from matches, People on a sample), not a way to run a whole library.

2. digiKam — best free, catalog-based

FREE · OPEN SOURCE (KDE) · WINDOWS 10+, MACOS, LINUX · VERSION 9.1.0 (JUNE 2026)

The most capable free photo manager, with deep-learning face detection and recognition (OpenCV DNN, GPU-accelerated) that runs locally, over albums mapped to disk folders and a SQLite or MariaDB database.

Reasons to choose it: free, three platforms, and face recognition, tags and ratings in one catalog.

Reasons to skip it: it is a database you look after, with a steep learning curve. Searching from a reference photo and exporting one person to a folder are not described on its features page, so we do not claim them.

3. Apple Photos People — best if your photos are already in Apple Photos

FREE · BUILT INTO MACOS · MAC ONLY

The People album groups faces with nothing to install. But Apple Photos is library-based: photos are kept inside a single Photos Library package, not your own Finder folders, so it only helps with photos imported into it.

Reasons to choose it: your photos already live in the Photos library and you only use a Mac.

Reasons to skip it: photos on external drives or a NAS must be imported first, and there is no Windows version.

4. Immich — best self-hosted, with phone access

FREE · OPEN SOURCE (AGPL-3.0) · SELF-HOSTED SERVER (DOCKER) · V3.2.2 (SEPTEMBER 2026)

The self-hosted Google Photos replacement most people land on: phone backup, facial recognition and clustering, CLIP and text-in-image search, on a server you run.

Reasons to choose it: you want your photos and faces on your phone, from anywhere, without Google.

Reasons to skip it: you become the administrator — Linux recommended, Docker and Compose, at least 6 GB of RAM (4 GB only with machine learning off), and a database to back up.

5. Mylio Photos — best for a library synced across devices

SUBSCRIPTION · $240/YEAR OR $25/MONTH (INDIVIDUAL) · MACOS, WINDOWS, IOS, ANDROID · 7-DAY TRIAL

A folder-preserving library with offline face recognition, AI search, duplicate review and peer-to-peer sync between your devices.

Reasons to choose it: the same library on phone and computer without a cloud service holding it.

Reasons to skip it: it is a subscription — one year costs more than most one-time options here.

6. Excire Foto 2027 — best local AI catalog

ONE-TIME €229 · MACOS, WINDOWS · ALL AI LOCAL

A catalog with on-device face recognition, automatic keywording and natural-language prompt search across content, faces, scenes and text; Excire says everything runs locally. For Lightroom Classic there is the Excire Search 2026 plugin (€199).

Reasons to choose it: a full local AI catalog with keywords and ratings.

Reasons to skip it: the price, and the catalog — it organizes inside its own collections and does not claim to sort files into folders.

7. Tonfotos — best for family archives

FREE VERSION + PERSONAL $59 / FAMILY $159 ONE-TIME · WINDOWS, MACOS, LINUX

On-device face recognition, a view of your photos by year and date without relocating the files, a duplicate finder and family-tree features.

Reasons to choose it: a family archive where faces connect to a family tree.

Reasons to skip it: the free version limits how many people it recognises and collects anonymous statistics.

8. RetinaTag — best no-cost newcomer

FREE · MACOS (APPLE SILICON), WINDOWS X64 · 100% OFFLINE

A free, offline app with on-device face clustering, local CLIP search and EXIF-date sorting when importing from USB or SD cards.

Reasons to choose it: you want to try offline face clustering for nothing.

Reasons to skip it: it is a very young one-developer project (repository created in March 2026), and it does not list Intel Macs.

9. PhotoPrism — best self-hosted alternative to Immich

FREE COMMUNITY EDITION (AGPL-3.0) + PAID MEMBERSHIPS · SELF-HOSTED SERVER (DOCKER)

A self-hosted library with face recognition, automatic labels, reverse geocoding and duplicate detection; it can also read metadata from Google Photos exports.

Reasons to choose it: a self-hosted web library, if you prefer it to Immich.

Reasons to skip it: it is a server — at least 2 cores, 3 GB of RAM and 4 GB of swap, Docker recommended.

How to find every photo of a person with FolioSort

About five minutes, plus a one-off indexing run.

  1. Install FolioSort and open Find a Person. Download FolioSort for Mac or Windows and open Find a Person in AI Search. No account is needed.
  2. Add the folders to search. Choose Add folders to index and include every drive and folder where their photos might be. Indexing happens once; after that, searching is immediate.
  3. Add 1–3 reference photos. Pick clear, front-facing photos. For a child, use photos from different ages: a three-year-old and a ten-year-old are hard to link from one face.
  4. Run Find This Person and review the matches. Matches arrive sorted by confidence. Use the match-strength slider to make it stricter or looser, and fix any photo that landed on the wrong person.
  5. Copy the matches into a folder. Select the keepers and choose Copy to Folder. Copies go to the folder you pick; every original stays where it was.

No reference photo? Open People instead and let it group the faces first. The full control-by-control guide is in the Find a Person help page.

What do people use it for?

Every photo of a child, for a slideshow. Add reference photos from a few different ages, copy the matches into one folder and sort it by date. A Smart Folder narrowed to that child keeps it up to date afterwards.

A wedding guest's photos. Find a Person pulls one guest out of the whole shoot; for a household or the wedding party, a group search returns photos of any member and can lay the copies out as a folder per person.

Someone who has passed away. When a family loses someone, the photos are needed quickly — for the service, a memorial slideshow, or to share with relatives — and they are scattered across decades of drives. Photos from different periods of their life make good references. Nothing is uploaded, and copying leaves the archive exactly as it was.

A school class or sports team. Name each child in People, then split the season's photos into one folder per child in one run; a photo with several of them is copied into each folder.

How we compared

Each rival was checked on its vendor's own site, documentation, official repository or store listing on 23 September 2026 — no third-party reviews, and anything not stated is marked “—” rather than guessed. FolioSort's facts come from the app and its own feature and help pages. We make FolioSort, which is why it is first.

Questions people actually ask

How do I find all photos of a person on my PC without Google Photos?

Use a desktop app that does face recognition on your own computer. In FolioSort, open Find a Person, add the folders to search and 1–3 clear photos of the person, and run the search; matches come back sorted by confidence and you copy the keepers into a folder. Nothing is uploaded and no account is needed, on Windows 10/11 or macOS 14+. For a free catalog instead, digiKam also does local face recognition.

Is there free offline face recognition software for photos?

Yes. digiKam is free and open source on Windows, macOS and Linux, with local face recognition inside its catalog. RetinaTag is free and offline on Apple Silicon Macs and Windows, but very young. Tonfotos has a free version that limits how many people it recognises. FolioSort lets you copy 20 photos from person matches for free before you decide on Elite.

Does Windows Photos or Apple Photos group photos by face?

Apple Photos has a People album, but only for photos imported into the Photos library, and only on a Mac. On Windows, Microsoft's own support page says face grouping is not in the new Photos app; the People tab exists only in Photos Legacy.

Are my photos or faces uploaded when I use FolioSort?

No. Face detection and recognition run on your own computer with models bundled in the app: Core ML on the Mac's Neural Engine, local ONNX models on Windows. There is no account and no server, it works offline, and the face index stays on your own disk. Clear index deletes the face data without touching any photo.

What is the difference between Find a Person and People?

Find a Person starts from a photo: add 1–3 reference photos and it finds every other photo of that person. People starts from nothing: it groups the faces in a folder into people on its own, and you name and merge them afterwards. Use the first when you know who you are looking for, the second to see who is in the library.

Can I use the names in Lightroom, Capture One or digiKam?

Yes. FolioSort can write the names you give people into XMP sidecar files next to each photo, as a flat keyword and under a People > Name hierarchy, which Capture One, Lightroom Classic and digiKam read as keywords. Original files, including RAW, are never opened for writing, and an existing sidecar is updated rather than replaced.

How much does FolioSort's face recognition cost?

Find a Person and People are part of FolioSort Elite, a one-time €39.99 for 3 devices across Mac and Windows, with no subscription. Before paying you can copy 20 photos from person matches and try People on a sample. There is a 14-day full refund with no conditions, even if the licence key has already been activated.

Find everyone who matters, without uploading a face

Free to download. Try Find a Person on your own family — 20 photos copied from matches on the free plan — then unlock it with Elite, a one-time €39.99 for 3 devices across Mac and Windows, with 14 days to change your mind. See pricing for what each plan includes.

macOS 14+ and Windows 10+ · No account · Nothing leaves your computer

Download Free