Add your photos

The finished example searches photos by what they look like — no tags, no captions. The trick is the model: an image embedding model reads each photo into the same "space" a query image lands in.

Create a list with an image model

Create a list — call it photos — and attach an image embedding model. Where a text model reads words into meaning, an image model reads pictures into it: two visually similar photos land close together, so "find ones like this" becomes a distance check. That's the one piece of setup that makes visual search possible.

Load the photos

Each photo is a record with an image field pointing at the picture — a URL, or a file in a media store. Import a folder of them the same way you'd import any records; there's nothing to tag. The model embeds each image as it lands, so the list becomes visually searchable on its own.

POST https://api.searchstack.dev/search-result/Demo/photos/with-fields
X-API-Key: {your key}
Content-Type: application/json

[
  { "name": "Harbour at dawn", "image": "https://.../harbour-dawn.jpg" },
  { "name": "Pine ridge",      "image": "https://.../pine-ridge.jpg" }
]

Go deeper: Embedding models, including image models in the reference.