> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pureframe.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Search signals

> Understand how visual and transcript signals affect search results.

Pureframe AI can match what appears on screen and what people say. Understanding these signals helps you choose a search mode and interpret results.

## Visual search

Visual search finds conceptually similar content. A query such as `whiteboard with a diagram` can match relevant frames even when those words are never spoken.

Use visual search when appearance, objects, settings, actions, or composition matter. You can search with text or a reference image.

## Transcript search

Transcript search finds spoken words. It works best when your query contains wording that is likely to appear in the video.

For example, `pricing objection` is more likely to match a segment containing those terms than a segment that expresses the same idea with unrelated words.

## Scene and on-screen text signals

* `scene` results describe the type of scene shown, such as a whiteboard or an outdoor setting.
* `ocr` results match text visible in the frame, such as slide content, signs, or captions.

Scene availability depends on your plan. On-screen text results can appear alongside other result types. See [Combined queries](https://docs.pureframe.ai/search/combined-queries.md) for mode behavior.

## Choose a search strategy

* Use `modes=video` when you care about what appears on screen.
* Use `modes=transcript` when you know what someone said.
* Use the default combined search when either signal could identify the moment.

Combined search returns one relevance-ranked result set. A segment can appear because it matches visually, through speech, or through both signals.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.