Skip to content
SPECIFICTION
Get listed

Facet

ncoevoet.github.io

Free plan Open source WebWindowsmacOSLinuxAPISelf-hosted
Visit ncoevoet.github.io
Spec score 8.9 #3 of 29 in AI Photo Culling Software

What it is

Facet is a free, local tool for scoring, culling, organizing, and serving a web gallery for photo libraries. It runs on the user's machine without cloud services, accounts, or API keys. Its scoring evaluates nine dimensions, including aesthetic quality, composition, sharpness, color, exposure, face quality, subject saliency, and dynamic range. It detects bursts, similar photos, blinks, and duplicates, with a dry-run preview for automatic culling. The gallery includes natural-language semantic search, filters, timeline and map views, albums, and themed slideshows. Face tools group people and support searching, renaming, merging, and organizing clusters. XMP sidecars provide metadata interoperability with Lightroom, Capture One, digiKam, and darktable. Users can send photos to a WebDAV inbox, or share revocable album links that need no recipient login; proofing supports hearts and comments, with an optional PIN. Core functions work on CPU, while an NVIDIA GPU unlocks additional models and features. Facet can be installed with Docker or natively on Linux and macOS, requires Python, and does not provide RAW editing or development.

Spec sheet

What it does
Facet analyzes local photo libraries, scores images, supports culling, and serves a browser gallery.Read 15 Oct 2026 · ncoevoet.github.io
Product
Facet scores, culls, organizes, and serves a web gallery for local photo libraries.Read 28 Sep 2026 · github.com
Audience
The maker describes Facet as a fit for people with large local libraries who want to find top shots and cull bursts and near-duplicates.Read 28 Sep 2026 · github.com
Integrations
XMP sidecars connect Facet with Lightroom, darktable, digiKam, Capture One, and Immich workflows.Read 15 Oct 2026 · github.com
Privacy
Facet runs on the user's machine and says it uses no cloud, accounts, or API keys.Read 28 Sep 2026 · github.com
License
Facet is distributed under the MIT license.Read 28 Sep 2026 · github.com
API extensibility
The documentation describes configurable plugins and a Prometheus metrics endpoint that can be enabled through configuration.Read 15 Oct 2026 · github.com
Burst culling
Facet groups rapid-fire shots and recommends keepers using quality, sharpness, and blink signals.Read 15 Oct 2026 · github.com
Community technical description
A Reddit project post describes Facet as a free open-source Python tool using multiple vision models, a web gallery, face clustering, semantic tagging, and GPU-profile detection.Read 15 Oct 2026 · reddit.com
Culling
It detects bursts, similar photos, blinks, and duplicates, and provides auto-culling with a dry-run preview.Read 28 Sep 2026 · github.com
Documentation depth
The project publishes separate guides for installation, commands, configuration, scoring, faces, viewer use, interoperability, Immich, and deployment.Read 15 Oct 2026 · github.com
Face recognition
The gallery detects faces, clusters people, supports naming and merging, and provides blink and expression indicators.Read 15 Oct 2026 · github.com
Gallery
The gallery includes natural-language semantic search, filters, timeline, map, albums, and themed slideshows.Read 28 Sep 2026 · github.com
Gallery views
The web interface includes mosaic and grid galleries, timeline, map, folders, albums, memories, slideshows, and themed capsules.Read 15 Oct 2026 · github.com
Hardware
Core scoring, face detection, culling, gallery, search, albums, and metadata export work on CPU without a GPU; an NVIDIA GPU unlocks additional models and features.Read 28 Sep 2026 · github.com
Hardware profiles
Profiles cover CPU-only legacy mode, NVIDIA GPUs with 8GB, 16GB, or 24GB VRAM, and automatic Apple Silicon sizing.Read 15 Oct 2026 · github.com
Independent directory description
Localhosted lists Facet as a self-hosted AI photo manager that runs on CPU or GPU without cloud services or accounts.Read 15 Oct 2026 · localhosted.com
Independent directory metadata
Localhosted reports Facet as MIT licensed, with Windows, macOS, Linux, Docker, and web availability, and recorded 288 repository stars when fetched on 15 October 2026.Read 15 Oct 2026 · localhosted.com
Installation
Docker is supported on Windows, macOS, and Linux, while native installation is documented for Linux and macOS.Read 15 Oct 2026 · github.com
Licence
The repository distributes Facet under the MIT License, permitting reuse, modification, distribution, sublicensing, and sale subject to the notice.Read 15 Oct 2026 · github.com
Lightroom plugin
The bundled Lightroom Classic plugin exchanges ratings, flags, metadata, burst decisions, and collection information through a manifest.Read 15 Oct 2026 · github.com
Limit: Lightroom picks
The Lightroom integration rejects discarded frames but does not yet mark the recommended burst frame as a Pick automatically.Read 15 Oct 2026 · github.com
Limit: Lightroom stacks
The maintainer explains that the Lightroom SDK cannot create stacks, so Facet cannot automatically build them and users must use Lightroom’s own auto-stack command.Read 15 Oct 2026 · github.com
Limit: no editing app
The project documentation positions Facet as a scoring, organization, and culling tool rather than a RAW development editor or mobile replacement.Read 15 Oct 2026 · github.com
Limit: setup reliability
A Reddit user said an initial trial could not make Facet see their photos, despite trying the project.Read 15 Oct 2026 · reddit.com
Limits
Facet does not provide RAW editing or development, and its README says it has no mobile app or cloud backup.Read 28 Sep 2026 · github.com
Maker identity
The repository owner identifies as Nicolas Coevoet.Read 15 Oct 2026 · github.com
Maker location
The maker’s GitHub profile gives the location as South of France.Read 15 Oct 2026 · github.com
Model download size
The first scan downloads approximately 4.7GB for legacy, 6.9GB for 8GB, 14.6GB for 16GB, or 19.1GB for 24GB profiles.Read 15 Oct 2026 · github.com
Notable release
Release v1.19.0, named Chambre noire, was published on 5 October 2026 and added manual tags, a gallery context menu, a scan button, and a multilingual getting-started guide.Read 15 Oct 2026 · github.com
Original-file safety
Scanning is read-only; metadata writing is a separate opt-in operation and RAW originals are not modified by embedding commands.Read 15 Oct 2026 · github.com
People
Facet detects and groups faces, with tools to search, rename, merge, and organize person clusters.Read 28 Sep 2026 · github.com
Personal ranking
A/B comparisons, ratings, favorites, and culling decisions train a personal My Taste ranking model.Read 15 Oct 2026 · github.com
Phone ingestion
A WebDAV inbox supports automatic uploads from PhotoSync or another WebDAV client.Read 15 Oct 2026 · ncoevoet.github.io
Phone uploads
Users can send photos to its built-in WebDAV inbox with PhotoSync or another WebDAV app.Read 28 Sep 2026 · github.com
Photo scoring
It scores photos across nine dimensions, including aesthetic quality, composition, face quality, sharpness, color, exposure, subject saliency, and dynamic range.Read 28 Sep 2026 · github.com
Praise: local workflow
A self-hosting Reddit commenter reported being super impressed after testing the scoring workflow on a large photo collection.Read 15 Oct 2026 · reddit.com
Praise: overall capability
A Reddit user described Facet as very impressive with substantial potential while testing it as an Aftershoot alternative.Read 15 Oct 2026 · github.com
Privacy model
The product says photos stay on the user’s machine and that it requires no cloud service, account, or subscription.Read 15 Oct 2026 · ncoevoet.github.io
Public API
The viewer exposes documented REST endpoints under paths such as /api/photos, /api/filter_options, /api/culling-group, and /api/export/sidecars.Read 15 Oct 2026 · github.com
Scoring dimensions
Images are scored for aesthetics, composition, face quality, eye sharpness, technical sharpness, color, exposure, subject saliency, and dynamic range.Read 15 Oct 2026 · github.com
Security behavior
The changelog says access-controlled installs now prevent anonymous metadata leakage and fail closed when configuration is corrupt.Read 15 Oct 2026 · github.com
Semantic search
Natural-language queries such as sunset on the beach search the library through image and text embeddings.Read 15 Oct 2026 · ncoevoet.github.io
Sequence protection
Exposure brackets, panorama sweeps, and HDR panoramas are recognized as sets and kept together during culling.Read 15 Oct 2026 · github.com
Setup
Facet can be installed with Docker or natively on Linux and macOS; its README says it requires Python.Read 28 Sep 2026 · github.com
Sharing
Albums can generate revocable links, and client proofing allows recipients to heart photos and leave comments.Read 15 Oct 2026 · github.com
Supported formats
Documented inputs include JPEG, HEIF/HEIC/HIF, ten RAW formats, PNG, GIF, WebP, BMP, TIFF, and AVIF.Read 15 Oct 2026 · github.com

Plans and prices

Facet Free Free and local · no cloud, accounts, or subscriptions Read 28 Sep 2026 · github.com
Free Free local processing · self-hosted gallery · no cloud account or subscription Read 15 Oct 2026 · ncoevoet.github.io

Every Facet plan and price →

Strong points and limits

Strong points

  • Runs locally without cloud, accounts, or API keys.
  • Detects bursts, blinks, and duplicates.
  • Gallery includes semantic search and map views.
  • Shares revocable album links without recipient login.
  • Core features work on CPU.

Limits

  • Does not provide RAW editing or development.
  • No mobile app or cloud backup.
  • Requires Python.

Who it suits

Facet suits people with large local photo libraries who want to locate strong shots and cull bursts or near-duplicates. It is a fit for users comfortable with local installation and without a need for RAW editing or cloud backup.

Verdict

Facet combines local photo discovery, culling, and gallery sharing while keeping core processing on the user's machine. Its lack of RAW development, mobile app, and cloud backup sets clear boundaries for its use.

Where Specifiction places it

Alternatives to FacetAll alternatives →