Semantic Image Search

Semantic search lets you find images by describing what you want instead of providing a reference image.

Steps

1. Click Semantic in the top-right toolbar. If unavailable, first add a folder and generate semantic data.
2. Enter a natural-language description.
3. Click the options button to select the entire database, a folder or a collection.
4. Set the maximum number of results.
5. Start the search and review the returned images.

The results open in the Semantic Search Results drawer, and the Results sidebar retains the search history. In the illustrated example, “白雪覆盖的群山” (snow-covered mountains) returns the mountain photo and its backup first.

Writing Better Prompts

Describe the subject, scene, color, style, and use case when possible. For example:

• "product photo on a blue background"
• "modern living room with warm lighting"
• "woman holding flowers outdoors"
• "minimal poster with red elements"

Short prompts are fine, but clearer descriptions usually make the first result set easier to review.

Search Options

Use scope to limit the search to a folder, collection, or database range. Use result count to control how many candidates return.

Common Cases

• If results are too broad, add subject, color, or scene details.
• If no results appear, check whether semantic data has been generated.
• If the first prompt is not enough, run a second search with more specific wording.
• Semantic prompts describe visual content; the words need not appear inside the image. To find actual text in images, generate OCR data and use text/OCR search or the toolbar’s OCR field.