Software · Self-hosted
Your own photo cloud — with Lightroom Classic instincts.
Where Lightroom chugs, photoArchive flies: browse terabytes of photos at lightning speed with tiered, tunable caching — hosted on your own computer, NAS, or server. Google Photos convenience, Lightroom Classic control, and your files never leave your network.
47,000+
Photos in the author's daily-driver archive
100%
Local — nothing leaves your network
3
Search engines fused per query
1
Consumer GPU runs all of it
From Google Photos
The instant timeline, semantic search, face grouping, shareable links, and a phone app that installs like the real thing.
From Lightroom Classic
The folder tree, pick/reject culling, stacks, keyboard-first loupe with lights-out, and filters that compose the way a working photographer thinks.
From neither
Elo photo ranking. Quick this-or-that picks teach the archive which photos are your best — no star-rating agony, just two-second decisions that compound.
The Library
Terabytes of photos on a sleepy external drive, browsed at lightning speed: a tiered preview cache with tunable budgets pre-generates in the background, an in-memory hot cache serves from RAM, and a virtualized grid keeps scrolling silk-smooth at any depth. A real folder tree, composable filters, and a timeline scrubber complete it. It feels like Lightroom Classic without the chug — because that was the bar.
Ranking, not rating
Forget agonizing over star ratings. Refine shows you a mosaic — you pick the best one. Behind every two-second decision is an Elo engine with uncertainty tracking that propagates results through visually similar photos. A quality meter tells you exactly how sorted any folder, shoot, or search is.
Search that understands
Every photo is indexed three ways on your own GPU: metadata full-text, semantic vision embeddings, and rich VLM-written captions. Queries fuse all three engines with reciprocal-rank fusion, and the omnibox streams live photo results, facet completions, and natural dates as you type. No cloud API. No subscription. Measured by a built-in eval harness.
Stacks & safe trash
Three automatic builders detect burst sequences, export variants of the same edit, and cross-source duplicates — like the compressed Facebook copy of your original — and stack them behind a single cover, Lightroom-style. Resolving a stack moves losers to a fully restorable on-drive trash. Nothing is ever deleted until you say so.
Share any collection as a password-protected gallery with view analytics and client proofing — their favorites flow back into your archive as picks. Public links serve previews only; originals never leave the machine.
A real PWA: fast timeline, pinch density, pull-to-refresh, haptics, one-handed action bars, and an Android Back button that behaves natively. Pair it with Tailscale and your archive is in your pocket on your own private network.
Local face detection clusters faces into people entirely on your hardware. Name them, merge them, filter any view by who is in the frame. Face data never leaves your network.
The app indexes your folders and builds its own caches — it never edits, moves, or deletes a source file. Offline drives degrade gracefully: cached views keep working until the drive returns.
Thumbnail generation, semantic embeddings, face scanning, and VLM captioning share a single consumer GPU through sequential coordination. The live archive runs on an RTX 2060 Super.
FastAPI + SQLite, browser-native ES modules with no build step, additive-only migrations, and a contract-tested API surface with 380+ tests. Built to still work in ten years.
Open source
photoArchive is built in the open and runs on hardware you already have. Point it at a folder of photos and it takes care of the rest.