The goal is to build a local self-hosted photo library with automatic face recognition, AI-generated captions, and natural language semantic search while preserving privacy.
Immich is widely recommended for its mature open-source stack with Docker support and built-in AI features.
Photoprism offers good face recognition but suffers from performance and contributor bottlenecks.
Ente provides end-to-end encrypted cloud or self-hosted storage with local AI processing.
Nextcloud with Recognize and Memories apps can be a viable alternative with Docker and cron-based batch pipelines.
Model recommendations include Mistral Small, Gemma3, Qwen via Ollama, SmolVLM, BLIP, SentenceTransformer, and InsightFace with OpenCV pipelines.
Semantic vector databases like ChromaDB and using CLIP embeddings can optimize search performance.
Docker containers on devices like Synology or VPS with S3-compatible object storage can host the entire stack.
Get notified when new stories are published for "General AI News"