Meme Search

Meme Search is an open source meme search engine and finder that helps you organize and retrieve images using AI-generated descriptions and semantic vector search. It performs image-to-text extraction, embedding generation, and searching locally, designed to run easily via Docker.

Key Features

• Local image-to-text processing with multiple selectable captioning models
• Automatic and manual meme description management to improve search quality
• Fast semantic and keyword search backed by PostgreSQL with pgvector
• Tagging and filtering by tags, directory paths, and search mode (keyword vs vector)
• Bulk description generation for faster indexing
• Directory rescans to detect and index newly added images
• Drag-and-drop uploads via the web UI (with configurable storage directory)
• Dark mode user interface

Use Cases

• Personal meme library management for quick semantic retrieval by content or text
• Homelab media organization with local-first AI indexing and search
• Team/shared collections where tags and edited descriptions improve discoverability

Limitations and Considerations

• First-time caption generation can be slow due to downloading and caching model weights
• Resource usage depends on the chosen image-to-text model size and hardware capabilities

Meme Search combines a Rails-based web interface with a Python-based generator to provide practical, private semantic search for image collections. It is well-suited for users who want fast local search, tagging, and ongoing rescans without relying on external AI services.