Semantic Search

Search by describing item features in natural language without needing the exact item name.

Plain explanation

Traditional search requires exact item names, such as “ZARA black suit pants”. Semantic search lets you describe items in everyday language, such as “that black pair of pants” or “winter shoes”. AI understands the appearance of your item images and finds items that match the description. It is useful when you cannot remember the exact name but remember what the item looks like.

What Is a “Search Index”?

Plain explanation

To let AI “find items by appearance”, AI first needs to scan all your item images and remember the visual features of each one. This process of “letting AI look once and remember” is called building a search index. After the index is built, future description-based searches can return results quickly.

Prerequisites

Semantic search requires a search index first:

  • Turn on “Enable Search Index” in the search dialog
  • The system builds an AI index for item images; the first build may take a few minutes depending on the number of items
  • Semantic search can be used after the index is built
  • New items are indexed automatically afterward, with no manual action needed
Member Feature

Search indexing requires membership access.

How to Use

  • Tap the search icon on the home page in the top toolbar
  • Enter a description of item features in the search dialog
  • Make sure “Enable Search Index” is turned on
  • Submit the search
  • View the list of matching items
Screenshot placeholder: search dialog and semantic search results

Search Examples

You can describe item features like:

  • “black pants”
  • “white T-shirt”
  • “bottled water”
  • “winter shoes”
  • “red coat”

Semantic Search vs Image Search vs Regular Search

Differences

Regular search: enter text keywords to match item names or note fields.
Semantic search: enter a text description and match against visual features from item images.
Image Search: take or upload an image directly to find similar items.

Which one should you use?

  • If you know the item name: regular search is fastest
  • If you remember only the appearance, not the name: use semantic search with a text description
  • If you have a real photo or online image and want to find similar items in your library: use image search

Use Cases

Scenario 1: Finding clothes in the wardrobe

Winter arrives and you want to find that camel-colored coat, but you forgot what you named it. Enter “camel long coat” in search, and semantic search can locate it directly.

Scenario 2: Quickly finding supplies in storage

If your storage contains many daily supplies and you want to find “that red bottle of shampoo”, describe it as “red bottle shampoo” and find it without checking items one by one.

Scenario 3: Helping family members find things

Older family members may not remember item names but can describe appearances. For example, “medicine in a white square box” can be used directly to locate the item.

Scenario 4: Finding collectibles by appearance

When you have hundreds of figures or models and want to find “the one holding a sword” or “the one with red armor”, semantic search can retrieve items by image features faster than browsing categories.

How the Search Index Works

The search index works like this:

  • The system analyzes visual features in item images
  • It builds a semantic index of image content
  • When you enter a description, it matches image features
  • It returns items whose visual features match the description

Notes

  • Your description should focus on the item’s visual features, such as color, shape, material, and style
  • Descriptions about time, status, or numeric attributes are not supported; for example, use filters for “bought last month” or “3 left”
  • The more specific the description, the more accurate the results
  • Items need images to be indexed; items without images cannot be found this way
Tip

Semantic search is best for finding items by visual features. If you need to filter by time, status, price, or other attributes, uselist filters.

Related Articles