Home

LrGeniusAI Wiki

Welcome to the project wiki.

---

User Documentation

Getting started

• Getting Started — install, download the models, analyze your first photos
• FAQ — common questions and quick answers
• Troubleshooting — connection issues, API errors, missing models
• Performance Tips — which features are LLM-costly and how to speed them up

Feature guides

• Help: Analyze and Index — AI tagging, descriptions, search embeddings, species identification
• Help: AI Edit Photos (beta) — generate and apply Lightroom develop recipes
• Help: Advanced Search — semantic free-text search
• Help: Cull Photos (beta) — burst grouping, ranking, Picks/Alternates/Rejects
• Help: People & Faces — face clusters, person names, face-based collections
• Help: Find Similar Images — near-duplicate and content-similar search
• Help: Keyword Deduplication and De-Clutter — synonym merging and auto de-clutter
• Help: Train from Edits — save your edits as AI style examples
• Help: AI Agents (MCP) — let Claude, LM Studio or another AI agent search your catalog and work in Lightroom
• Plugin Guide — complete menu reference and workflow overview

AI model setup

• Help: Choosing AI Model — cloud vs local, model comparison table
• Help: Local AI Models — built-in llama.cpp & MLX engines, no external app
• Help: Ollama Setup
• Help: LM Studio Setup
• Help: Other AI Server — OpenRouter, llama.cpp server, LiteLLM, vLLM
• Help: Docker with GPU Access — the backend in a container on another machine, with a local AI model on its NVIDIA, AMD or Intel GPU

Other

• Disclaimer — no warranty, AI results, cloud providers, faces, AI agents
• Credits & Dependencies

---

Developer Documentation

• Build Environment Setup — Windows & macOS toolchain setup for server-rs, including the llamacpp feature and MLX sidecar
• Backend API Reference — all REST endpoints documented
• Dev: Agent Bridge and MCP Endpoint — how agents drive Lightroom, the command contract, and testing the plugin through MCP
• Dev: Agent Chat — the in-app chat: a local model works the catalog through the MCP tools, with the user's approval
• Server Guide — backend architecture, database backup, lifecycle
• Dev: Feature Priority Decision
• Dev: Image Culling Implementation Plan
• Dev: Culling — what the other tools judge — Lightroom, Aftershoot, Narrative, FilterPixel, Imagen, Excire, Evoto
• Dev: Culling — the public benchmark — defects made on purpose on free-licence photos, and what it measured
• Dev: Capture One Port — Findings and Plan — what Capture One can and cannot do; investigated 2026-09-10, not pursued
• Dev: AI Edit via XMP — Findings and Experiments — what crs: XMP can express, mask coordinate frames, SDK apply paths, and the in-Lightroom experiments that settle the rest; researched 2026-09-26
• Dev: Native Develop Model — lrg-develop: the key registry, the typed develop model, the policy filter and the Lua/JSON reader; how to add a key

Auto-generated from README files

• Project README
• Plugin README
• Server README

---

What is LrGeniusAI?

LrGeniusAI is an AI extension for Lightroom Classic. It runs a local backend server and connects it to the Lightroom plugin to provide:

• AI metadata generation — keywords, titles, captions, alt text, via cloud APIs or models the backend runs locally itself
• AI develop edits (beta) — an AI retoucher that edits, looks at Lightroom's render and corrects it, with looks, your own instructions and AI masks; or recipes interpolated from your own saved edits, with no LLM involved
• Semantic free-text search — find photos by describing them in natural language
• Species identification — name the animals, plants and fungi in your photos down to species, on your own machine, with no cloud account
• Image culling (beta) — burst grouping, scoring, Picks/Alternates/Rejects collections
• Face & person workflows — face detection, clustering, named person collections
• Find similar images — near-duplicate and visually similar search
• Keyword management — automatic de-clutter and interactive synonym deduplication
• AI agents (MCP) — Claude, LM Studio and other AI apps can search your catalog and work in Lightroom, with every change undoable
• Style training — save your own edits as the examples AI develop edits are built from

For project overview and release info, see Project README.