Local AI Culling

Input—per 1M tokens

Output—per 1M tokens

Context—tokens

WeightsClosed

About

Local AI Culling is ranked #3 of 29 in AI photo culling software on Inferse. It runs on Linux, macOS, Windows.

Compared on AI photo culling software

Free plan: Yesgithub.com
AI quality scoring: Yesgithub.com
Subject detection: Yesgithub.com
Duplicate detection: Yesgithub.com
RAW photo support: Yesgithub.com
Blur detection: Yesgithub.com
Closed-eye detection: Yesgithub.com
Lightroom export: Nogithub.com

Facts

Purpose: Local AI Culling is an open-source, offline AI-assisted image culling application built for professional photographers.github.com · 30 Sept 2026
Workflow: It analyzes photo shoots and organizes images into KEEP, REVIEW, and REJECT folders for editing in Lightroom or another editor.github.com · 30 Sept 2026
Duplicate detection: It groups visually identical photos taken moments apart.github.com · 30 Sept 2026
Burst ranking: It identifies high-speed bursts and selects the best frame in each burst.github.com · 30 Sept 2026
Image evaluation: Its scoring evaluates technical quality including focus, noise, composition, expressions, and editability.github.com · 30 Sept 2026
Feedback dashboard: A local web dashboard lets photographers review decisions, adjust thresholds, and see why the AI made a decision.github.com · 30 Sept 2026
Export: The default hardlink export organizes files without duplicating or modifying the originals; copy mode is also configurable.github.com · 30 Sept 2026
Privacy: The security policy says images are not uploaded, cloud APIs are not used for image processing or metadata analysis, and usage telemetry is not collected or transmitted.github.com · 30 Sept 2026
Platforms: The README lists Windows, macOS, and Linux support, with hardware acceleration where available and CPU fallback.github.com · 30 Sept 2026
Installation: Installation requires Python 3.10 or later, with Python 3.11 recommended, and an SSD is highly recommended.github.com · 30 Sept 2026
Image formats: Documented supported formats include JPG, JPEG, PNG, and standard CR2 and NEF RAW; CR3, compressed ARW, and HEIC may have incomplete support.github.com · 30 Sept 2026
Dataset size: The project recommends 500 to 2,000 images per run and warns that datasets over 5,000 images may cause memory pressure and slower duplicate processing.github.com · 30 Sept 2026
Integration limits: The documented v1.0 limitations say Lightroom Classic plugin integration and video culling are postponed, and the app does not edit Lightroom catalogs or generate XMP sidecars.github.com · 30 Sept 2026
Support: The project directs users to its documentation and GitHub Issues for help, bug reports, and feature requests.github.com · 30 Sept 2026
Intended users: The application is built specifically for professional photographers.github.com · 30 Sept 2026
Culling features: It detects duplicates and bursts, ranks burst frames, and assesses focus, noise, composition, and facial expressions.github.com · 30 Sept 2026
Supported systems: The README lists Windows, macOS, and Linux support, with platform-dependent CUDA, Metal/MPS, or CPU acceleration.github.com · 30 Sept 2026
Editor workflow: The README says organized images can be edited in Adobe Lightroom or a preferred editor, while the limitations page says Lightroom Classic plugin integration is postponed.github.com · 30 Sept 2026
Image-format limits: Supported formats include JPG, JPEG, PNG, CR2, and NEF; CR3, compressed ARW, and HEIC may have incomplete support or metadata extraction.github.com · 30 Sept 2026
Dataset limits: The project recommends 500–2,000 images per run and warns that datasets over 5,000 images may use substantially more memory and process duplicates more slowly.github.com · 30 Sept 2026
Other limitations: The project says it does not edit Lightroom catalogs or generate XMP ratings, and video culling is postponed.github.com · 30 Sept 2026
License and fees: The repository identifies the project as open source under the MIT License and says local operation eliminates subscription fees.github.com · 30 Sept 2026