Is Best Photo Picker free?
Yes. It is MIT licensed with no account and no paid tier, and the Mac app runs the same code as the command-line install.
Point Best Photo Picker at years of photos and it scores every frame for sharpness, exposure and faces, then hands back the best fifty of one person. Near-duplicates are folded out and the scoring runs on your own CPU.
Compare Best Photo Picker with other photo apps. Add or remove apps to see how they differ.
| Requirement | Best Photo Picker | |
|---|---|---|
| Price | Free | Free |
| Licence | MIT | MIT |
| Web UI | Yes | Yes |
| Minimum hardware | Not stated | 2 GB RAM |
| Docker Containers | 1 (app only) | 2 (app + Writer) |
| Database | SQLite | SQLite |
| Architecture | x86 + ARM | x86 + ARM |
| Tech stack | Python | Go |
| Docker | Yes | Yes |
| Linux package | Yes | – |
| macOS | Yes | Yes |
| macOS | Yes | Yes |
| Linux | – | Yes |
Official client for desktop.
Yes. It is MIT licensed with no account and no paid tier, and the Mac app runs the same code as the command-line install.
No. Scoring, face clustering and the database stay on the machine. It downloads its AI models once (about 660 MB), checks for a new release, and loads OpenStreetMap tiles when a map is open.
No. It picks and copies a selection; keep the library app for storage, albums and sharing. The picks land in their own folder as ordinary files, and a hardlink or symlink export keeps the disk cost down.
Yes, headless. Mount the photo folder read-only and give it an output folder, and it copies the picks out. The full web app in Docker is published to the server's loopback only; from another computer, open it through an SSH tunnel.
Apple silicon, M1 or later. On an Intel Mac, Linux or Windows, install the PyPI package with Python 3.11 and run the web app from there.
Bring Apple Photos, Google Photos and camera files into one dated library on your own disk, then review each calendar date across every year.
Your phone, the camera and a drawer of old drives can all empty into one timeline that lives on your server. Frameleaf keeps the originals as ordinary files, recognises faces and pets on the machine, and opens the same library in the phone apps, the browser and the TV.
Ditch Google Photos for a FOSS library with automatic phone backup and facial recognition, all kept on your own hardware.
Manage your photo library with machine learning, from automatic face recognition and object detection to semantic search.
Import a shoot from the command line and Lychee files it into an album you tag, describe and share by public link, with a password if you want one. The self-hosted photo gallery keeps its own folders and database, shows the EXIF and IPTC data its photos carry, and browses as a timeline, a flow or a map.
Browse, organise and share photos and videos privately, with automatic labels and face recognition.
Point Photoview at the photo and video folders already on your server and browse them in a gallery whose albums mirror the directory tree, with thumbnails, EXIF data and RAW support. Face grouping is an optional feature you switch on at install, and it sorts people automatically. An album or a single photo can be shared by link, with a password if you want one.
Turn folders of photos and videos into a self-hosted gallery you scroll like a social feed, with reels, stories and a Places view built from the GPS tags already in the files. Originals stay where they are, a SQLite index and generated thumbnails keep it quick, and one Docker container serves the gallery to the household.