The Mac app that finds what’s in it
A local vision model writes a caption and tags for every image and video on your Mac — searchable in milliseconds. Nothing leaves the machine.
Library
receipt from the bike shop
Search by what’s visible.
Completely local
Milliseconds, not minutes
“I searched ‘whiteboard from the sprint’ and it found a photo I never named. That is the whole product.”
“Nothing leaves the machine. I finally have AI search I can run on client footage without uploading a byte.”
“Spotlight knows filenames. Lens knows the receipt, the beach, the Slack screenshot with the hex code. I stopped scrolling Photos.”
“The CLI captioned 4,000 files overnight. The app just opened and the library was already there.”
Search that sees
what you see
Lens waits until your library is captioned, then finds files by the scene, the text on screen, or a tag — not the filename.
8 results 2 ms
receipt from the bike shop
Searching… 1–10 ms
Natural language
Ask for a receipt, a beach, or the Slack screenshot with the hex code
Menu bar control
Live index status and a search field that stays out of the way
Ranked results
Name, path, caption, and recency scored together in a few milliseconds
Automatically understands
Captions
Tags
On-screen text
Filenames
Recency
Nothing leaves
the machine
Photos, videos, prompts, and captions stay on your Mac. Inference runs on a private loopback runtime — no account, no upload, no API key.
Learn More
On this Mac
Photos and videos,
one library
A Photos-style justified grid for stills, and video tiles that crossfade through frames on hover. Every file keeps its real aspect ratio.
Placeholder
Justified photo grid
Placeholder
Video hover frames
Fits your workflow
like a glove
Scan, thumbnail, and caption from the app or the CLI. The database is the contract — close the window and the job keeps going.
Captioning
CLI
App
Both
Placeholder
Choose a local vision model or keep the default
Filter by kind, tag, folder, date, or size
Add a caption lock so hand edits are never overwritten
Integrate search into your workflow with a scriptable CLI
lens scan
lens thumbs
lens caption
lens search
Run scan, thumbnails, and captioning from Terminal or the app
Designed for
The Mac
Lens is a fully native SwiftUI app with a Photos-style grid, Full Disk Access for the real library, and a Metal-accelerated local model runtime.
Download for MacRecent updates
Placeholder
Lens 0.1 — Local models, no cloud
A managed llama-server runtime, verified model downloads, and a Swift Models Library that can start and stop inference on this Mac.
Read more
Placeholder
Justified grid and video hover frames
The library now lays out like Photos, and video tiles crossfade through contact-sheet frames without regenerating anything.
Read more
Placeholder
Millisecond search over captions
FTS5 recall plus an fzf-style ranker. “receipt from the bike shop” is a real query, not a demo script.
Read moreGuides for Mac users
Placeholder
How local search actually works
Scan, thumbnails, captions, then a two-stage ranker. Why the database is the contract between the CLI and the app.
Read more
Placeholder
Why nothing leaves the machine
Loopback-only inference, hashed model downloads, and the security invariants Lens will not trade for convenience.
Read more
Placeholder
Getting started with the CLI
doctor, scan, thumbs, caption, search. The five commands that turn a home folder into a searchable library.
Read more