AI Photo Summaries

Fotohordr can generate a short natural-language summary for each photo — useful for semantic search (see Search) and for skimming a library without opening every image. This is entirely optional and off by default.

Turn it on from Settings → Analysis → AI:

Choosing a provider

• Local (on-device) — runs an embedding/vision model on your own hardware. Nothing leaves your machine. This is the default for new installs.
• Ollama — an already-running local Ollama instance, if you’d rather manage the model yourself.
• Cloud providers (Anthropic, DeepSeek) — send photos to a third-party API for summarization. Requires an API key. Faster and often higher quality, at the cost of photos leaving your machine for processing.

Summaries are generated as part of the normal background pipeline, the same way previews and face detection are — turning the feature on processes your existing library over time rather than requiring a separate batch step, and new photos are summarized as they’re imported.

Summaries generated on a host sync to every paired client automatically, the same as any other library metadata.