★ Star on GitHub

Get started

Install LUMORA. Keep every photo on your machine.

The open-source alternative to Mylio Photos — download a desktop build, import a folder, and search offline. No account. No upload.

Setup path

From download to AI search — all on your machine.

  1. Download
  2. Install
  3. Import Photos
  4. Index Library
  5. Search
  6. Enable AI Features

1. Install

Pick a release build, or compile if you’re developing.

Release installer

Fastest path for everyday use.

  1. Open the latest GitHub Release (or use Download to auto-pick your OS).
  2. Choose your platform:
    • macOSLUMORA_*_aarch64.dmg (Apple silicon)
    • Windows*_x64-setup.exe or .msi
    • Linux.AppImage, .deb, or .rpm
  3. Install, then launch LUMORA from Applications / Start Menu / your AppImage.

On Mac, if Gatekeeper blocks the app: System Settings → Privacy & Security → Open Anyway.

Build from source

For contributors and anyone who wants a local debug build.

Need: Bun, Rust (stable), macOS Xcode CLT. Optional: ffmpeg, jpegtran.

bun install
bun run tauri dev
bun run tauri build
  • Put ~/.cargo/bin on your PATH.
  • Dev UI also serves at http://localhost:1420.
  • See CONTRIBUTING.md.

2. First five minutes

One folder is enough to feel the product.

  1. Import photos Home → Import photos, or drop a folder on the window. Originals stay where they are; LUMORA only indexes.
  2. Browse the library Use the grid, Timeline, Albums, and smart collections (Videos, RAW, Screenshots, Documents…).
  3. Try search Filename and metadata work immediately. Filters like camera:iphone or rating>3 help narrow results.
  4. Optional: on-device AI Settings → AI — install CLIP for “dog on beach” search, OCR for text in images, faces for People. Downloads are checksum-verified and stay offline after install.
  5. Optional: watched folders Keep a camera roll or Downloads folder in sync so new files appear automatically.

3. Everyday usage

The loops you’ll use most. Deep detail lives in the full guide.

Search

Toolbar search mixes full-text and (when installed) semantic results. Recent queries show up as hints and under Discover.

Organise

Albums, tags, favourites, 1–5 ratings, colour labels. Multi-select from the grid; edit in the viewer.

Clean up

Duplicates sidebar: exact (bulk keep-one), near (review groups), blurry (Laplacian). Everything soft-deletes first — undo with ⌘Z / Ctrl+Z.

People & Places

Face clusters become searchable names. Places uses GPS + offline GeoNames — no map tile network.

Vault

Move sensitive shots into the Locked folder (Argon2id + XChaCha20-Poly1305). Store the recovery code somewhere safe.

Privacy

No telemetry. Network only for models you choose and optional update checks. Logs stay on disk.

System requirements

Recommended resources by feature.

Feature Recommended
Basic library Modern CPU
OCR 8GB RAM
Face recognition 16GB RAM
AI search (CLIP) GPU optional

Troubleshooting

AI models not downloading

You need a network connection only for the download. Check Settings → AI, retry the model, and confirm disk space. Downloads are SHA-256 verified — a failed verify means retry, not a partial install.

Slow indexing

First import of a large tree is the slowest. Leave the app open; indexing is incremental. Watched folders only process changes after the initial pass.

Missing thumbnails

Photos re-thumb on demand. For videos, install system ffmpeg so frame thumbs can generate; without it, video thumbs may stay blank.

Large libraries

Use a virtualised grid and keep originals on a fast disk when possible. Enable only the AI models you need — embeddings and faces cost CPU/RAM on first run.

Search not returning results

Confirm the library finished indexing. Filename/metadata search works without models; natural-language text-to-image needs CLIP installed and embeddings generated under Settings → AI.

Handy shortcuts

Full list in the guide.

Key Action
Space Toggle / close viewer
← / → Previous / next
F Favourite
0–5 Rate
Delete Soft-delete (restore from Trash)
⌘Z / Ctrl+Z Undo
Esc Close viewer / clear selection

Need more depth?

The full guide covers AI backends, duplicate edge cases, Places, vault recovery, and privacy defaults.