More than 60 formats stay in ordinary folders
Lap treats your existing directories as the photo library and supports more than 60 photo, RAW, and video formats. It handles Apple Live Photos, Google Motion Photos, and RAW plus JPEG or HEIC pairs as linked assets. Browsing can pivot by date, folder, map location, camera, lens, tags, ratings, faces, or saved Smart Album rules. That is a broad desktop workflow without first copying every original into a managed cloud.
The folder-first promise has an important limit. EXIF dates, camera details, GPS, and orientation come from the files, while collections, tags, comments, favorites, ratings, culling states, face data, and AI indexes stay in Lap's local database. Copying a photo outside the app does not carry that organization with it. Issue 40 asks for IPTC and XMP tag synchronization and was still open on September 23, 2026. Back up the database and configuration with the originals if that organization matters.
Local AI avoids photo uploads but still creates a second library
Lap runs text search, visual similarity, smart tags, face detection, and face clustering locally. The README claims optional multilingual search across more than 50 languages. Thumbnails, previews, embeddings, faces, and video indexes also remain on the device. Those derived files make search possible, but they create another body of data that needs disk space and a backup plan. Deleting the catalog leaves the original photos untouched while removing the work stored only in Lap.
Privacy is narrower than complete offline operation. The May 2, 2026 policy says photos, filenames, searches, tags, and face data are not uploaded to a Lap service. Map view can request OpenStreetMap tiles, update checks reach GitHub, and release builds enable Aptabase by default. Aptabase receives 2 lifecycle events, app started and app exited, with the Lap version, platform, and operating system. Buyers with a strict no-network rule should account for those documented connections.
What happened when we ran it
Our sandbox installed 673 Rust packages in 44 seconds. The checkout at commit 47ea34b contained 2,132 files, roughly 475,237 lines of source, and occupied 56.7 MB before the dependency work. Five CI workflow files were present, with no Dockerfile and no tests directory. The project under test lived in src-tauri/, which matches the Rust and Tauri core described by the README.
The build reached our 900-second limit without completing. A separate test command ran for 623 seconds and then failed with exit code 101. The supplied log tail shows CMake checks for ARM32, ARM64, pthread support, a potentially evaluated expression, and JPEG ICC profile support. Some probes passed and some failed, but the tail does not contain the final error. We cannot honestly assign a cause from that evidence. The useful conclusion is that commit 47ea34b did not complete our fresh Debian verification.
Building needs Node 20, Rust, models, and an FFmpeg sidecar
Source setup starts with Node.js 20 or newer, pnpm, stable Rust, and Tauri CLI 2. Platform libraries add Xcode tools on macOS or WebKitGTK, AppIndicator, SVG, Clang, NASM, CMake, and related packages on Linux. Two scripts download AI models and an FFmpeg sidecar before cargo tauri dev. That is a much larger path than downloading a release asset.
Packaged users have 8 downloads in the v0.3.2 release table across macOS, Windows, and Linux architectures. The macOS builds are notarized. Windows x64 and ARM64 MSI files are unsigned, so the README instructs users to bypass SmartScreen with Keep anyway. Linux gets DEB and AppImage choices, but H.264 playback depends on GStreamer and may require two extra plugin packages on Ubuntu, Debian, or Linux Mint. Platform coverage is wide, yet each platform keeps a specific installation caveat.
The v0.3.2 release is active while catalog gaps remain
Version 0.3.2 shipped on September 12, 2026, with map browsing, Motion Photos, date-based import, duplicate cleanup, and Linux AppImage fixes. The repository was pushed again on September 24. GitHub listed 2,719 stars and 38 combined open issues and pull requests when fetched, and several issues received updates that same day. This is current development activity, not a dormant release page.
Current reports still touch everyday library work. Issue 325 says deleting one image on a macOS 27.2 beta system temporarily blanked the whole library until scrolling. Issue 331 asks Lap to prefer JPEG previews for RAW pairs, and issue 40 covers portable metadata. None of those reports proves broad data loss, and the deletion report names a beta operating system. They do show where to run a trial: copy a representative folder, test deletes and rescans, inspect RAW pair behavior, then verify the catalog backup before pointing Lap at the only copy of a family archive.