AI Photo Search · Discovery

Find any photo by describing it — on your own computer

Type what you remember — "kids at the beach", "beach sunset in Faro", the words on a receipt, a place and a month — and FolioSort surfaces the matches, ranked, from your local library. Semantic meaning, text inside images, automatic events and recovered dates. It all runs on your device: no cloud, no account, no photo ever uploaded.

On-device MobileCLIP — no cloud, no account Same power as $149+ desktop AI tools, on-device
FolioSort AI Photo Search — natural-language results ranked from a local photo library

You remember the photo. You just can't find it.

Somewhere in 40,000 files is the beach evening in Faro, the receipt you photographed for the warranty, the whiteboard from that meeting. But nothing is named for it, and scrolling for twenty minutes usually ends with you giving up.

The tools that can search by meaning want your whole library in the cloud and your photos run through someone else's AI — turning the pictures you'd never post into training data and account rows.

Describe it in plain words — and it never leaves your Mac.

FolioSort's AI Photo Search understands what's in your photos. Type a description and an on-device MobileCLIP model ranks your library by meaning; ask for a word and it reads the text printed inside images. It even proposes trips and events for you, and recovers dates that EXIF lost.

Every model, every index and every query runs locally. There's no upload, no account and no telemetry about what's in your photos — the same on-device AI that desktop tools charge $149–189 for, folded into the app that already organizes your library.

What's inside

Five ways to find any photo

Semantic photo searchType natural language — "kids at the beach", "beach sunset in Faro" — and get ranked results by meaning. An on-device MobileCLIP model does the matching; no photo ever leaves your computer.
Text (OCR) searchFind any photo with readable text — receipts, documents, whiteboards, screenshots — by searching that text. Supports en/pt/es/fr/de and is accent-insensitive, so "fatura" finds "Fatura".
Auto-EventsFolioSort proposes trip and event clusters — "Faro — August 2019 · 312 photos" — from time and GPS. One click turns a cluster into an editable event you can rename, split or organize.
Date inferenceScans, WhatsApp images and exports with no EXIF date get a confidence-scored estimate from filename and context — so they stop landing in an /Unknown/ pile and slot into the right year and month.
Conversational search (beta · add-on)Ask compound questions — "screenshots with an order confirmation", "photos with a QR code" — by connecting FolioSort to a local LLM you run yourself. The model only ever sees your typed question, never your photos.
100% on-deviceEvery model, index and query runs on your Mac or PC. No cloud, no account, no telemetry about your photos' contents — and it all works offline.

How it works

From a vague memory to the exact photo

STEP 01

Point at a folder

Choose any folder — or your whole library — and FolioSort builds a local search index on your device.

STEP 02

Describe it

Type what you remember: a scene, a place and a month, or a word printed inside the photo.

STEP 03

See it ranked

Semantic and text matches come back ordered by relevance — the closest photos first.

STEP 04

Act on it

Open, or turn a result set or an auto-event into a folder — copy or move the keepers, always reversible.

Conversational search · beta · optional add-on

Ask a real question — with an AI that runs on your machine.

Compound questions are where keyword search gives up: "screenshots with an order confirmation", "photos with a QR code", "beach photos from August 2019". Conversational search handles them by connecting FolioSort to a local LLM you run — Ollama or LM Studio — on your own computer.

Your photos are never part of the conversation.

FolioSort translates your question into on-device searches over your local index. The language model only ever receives your typed question as text — never an image, never a thumbnail, never your library. Nothing is sent to a remote service.

It's a paid optional extra on top of Pro, and it's in beta — bring your own local model and it plugs straight in.

Why it's private

The AI comes to your photos — your photos never leave

On-device modelsMobileCLIP for meaning and local OCR for text run on your computer — not on a server.
Local indexThe search index and date inference are built and stored on your device, and work fully offline.
No account, no telemetryNothing about what's in your photos is uploaded or logged. There's no sign-in.
The LLM sees text onlyConversational search sends your question to a local model you run — never a photo.
Non-destructiveActing on results copies or moves files; Move goes to a folder or Trash and is undoable.
macOS & WindowsThe same on-device Discovery features ship on both platforms.
~94%
top-10 recall on our semantic-search test set
0 uploads
no photo ever leaves your device
$149–189
what rival desktop AI tools charge — you pay a fraction, once
5 languages
text search in en/pt/es/fr/de, accent-insensitive
I typed "beach sunset in Faro" and it pulled the exact evening out of eleven years of photos — no albums, no tags, nothing named. I'd looked at Excire, but paying $180 and importing everything into another app was a non-starter. This just searches the folders I already have, on my own laptop.
HOBBYIST / 60,000 PHOTOS
Describe. Rank. Keep.

FolioSort's on-device Discovery runs entirely on your device. AI Photo Search finds any photo by meaning or text; Find a Person pulls every shot of someone out of the pile; Find Screenshots sweeps out what isn't really a photo.

FAQ

AI Photo Search questions, answered

Are my photos uploaded or sent to any AI service?

No. Semantic search runs an on-device MobileCLIP model, text search uses on-device OCR, and the whole index is built and stored locally. No cloud, no account, and it works offline — no photo ever leaves your computer.

How is this different from keyword or filename search?

You don't need tags or filenames. Semantic search matches the meaning of what you type — "kids at the beach" finds the photos even if nothing is named that way. Text search reads the words printed inside an image, so you can find a receipt by what it says.

What is conversational search, and does it send my photos to an LLM?

Conversational search is an optional paid add-on (beta) that lets you ask compound questions in plain language. It connects FolioSort to a local LLM you run yourself (Ollama or LM Studio). The model only ever sees your typed question as text — never your photos. Nothing is sent to a remote service.

What languages does text (OCR) search support?

English, Portuguese, Spanish, French and German. Matching is accent-insensitive, so "fatura" finds "Fatura".

How does this compare to Excire Foto and other desktop AI photo tools?

Local AI photo tools such as Excire Foto typically cost $149–189. FolioSort does on-device semantic and text search too — for a one-time purchase that's a fraction of that — and folds it into the same app that organizes, renames and de-duplicates your library.

Is it free?

You can try AI Photo Search free. A one-time Pro purchase removes the free-plan limits, and conversational search is an optional paid add-on on top of Pro.

Stop scrolling. Start describing.

Point FolioSort at your photos and find them the way you remember them — by what's in them, the words on them, the trip you took — with every model and every query running on your own computer.

Free to try No account needed macOS 14+ & Windows 10+
Download Free