# How to find photos by the text in them (OCR)

> Open AI → Text in photos and click Read text. Windows' built-in text recognition reads signs, labels and documents on your PC; then type any word into search.

Source: https://firelet.media/help/search-text-in-photos
Published: 2026-10-05 · Updated: 2026-10-05
Applies to: Windows, macOS, Linux

To search text in photos with firelet.media Gallery, open the **AI** panel ([[Alt]]+[[3]]), expand **Text in photos** and click **Read text**. The gallery uses the text recognition built into Windows to read signs, labels, receipts, race numbers and documents in your pictures, all on your PC. Then type any word you remember into the ordinary search box and the matching photos appear.

## What Read text does

Plenty of photos are really about the words in them: a receipt you photographed for a claim, the sign at a trailhead, a whiteboard after a meeting, a screenshot of a booking. **Read text** looks at each photograph once, recognises the printed text in it, and stores those words in your library so the search box can find them.

The panel describes it like this: "Reads signs, labels, race numbers and documents with the text recognition built into Windows — nothing is downloaded or sent. Then the search box finds them by what they say."

- It reads photos, not videos.
- Online-only cloud files are skipped rather than downloaded.
- Your files are only read. The recognised text lives in the library, never in the picture.

## Steps: Read text

1. Press [[Alt]]+[[3]] to open **AI**.
2. Expand **Text in photos**.
3. Click **Read text**.
4. Watch the status count up ("Reading… 120 of 2,400"), then settle on a summary such as "2,400 of 2,400 read · text found in 318".

![The Text in photos section of the AI panel, with its description and the Read text button](https://firelet.media/help/media/shot-ai-text-in-photos.webp "Read text uses the text recognition built into Windows; nothing is downloaded")

To stop early, click **Stop** in the status bar: "Stopped after 120; what was read is kept." Clicking **Read text** again carries on with the photographs that haven't been read, and after you add new pictures it reads only those. When there is nothing left it says "Every photograph has been read."

> [!NOTE]
> Reading text adds to your library, so it pauses when the 14-day trial ends without a licence. Text already read stays searchable.

## Search text in photos by what the picture says

Use the search box on the toolbar above the grid, not the description box in the AI panel.

- Type one or more words from the picture, such as `invoice` or `platform 9`. Every word must match somewhere: in the recognised text, or in the name, folder, keywords, people, albums, places or titles.
- Matching ignores capitals.
- Add a date range or a folder filter to narrow a common word down.

To see what was recognised in a particular photo, open it and press [[I]]: the info panel includes the text found in the picture.

> [!TIP]
> Text search and AI description search work well together. Search for the words when you remember them ("gate B12"), and describe the scene when you don't ("airport departure board"). [How to find a photo by describing it](https://firelet.media/help/find-photos-by-description) covers the second kind.

## Filter: Text in the picture

To see every photo that contains text, or every photo that doesn't:

1. Press [[Alt]]+[[2]] to open **Filters**.
2. Expand **Quality & faces**.
3. Set **Text in the picture** to **Has text** (or **No text**).

This is a quick way to gather screenshots, receipts and scanned documents in one place. In **Advanced query** ([[Alt]]+[[9]]), the clause `has=text` does the same, and `has!=text` finds pictures with none. [How to sort and filter your photo library](https://firelet.media/help/sort-and-filter-photos) explains how filters combine.

## Languages

| Build | Engine | Languages |
|---|---|---|
| Windows | Text recognition built into Windows | The first language in your Windows language list that has text recognition installed, falling back to English |
| macOS and Linux (being finished) | Tesseract, after **Download English text model** (about 2.9 MB) | English |
| Android companion (coming soon) | None | No text reading |

On Windows nothing extra is downloaded: the gallery uses the recognition that comes with your Windows language packs. If none is installed, the panel says "Windows has no text-recognition language installed. Add one under Settings › Time & language › Language." Recognition is meant for printed text, such as signs, labels, race numbers and documents.

## Privacy: nothing is downloaded or sent

On Windows, reading text happens entirely on your PC with the recognition Windows already has. No model is downloaded, no picture is uploaded, and the words it finds are stored in your library database next to your ratings and albums. That includes anything sensitive in your photos, such as account numbers on a receipt: it never leaves your computer.

The [what firelet.media Gallery sends](https://firelet.media/help/what-data-is-sent) page lists the few connections the app does make, and none of them involve your pictures. For ideas on what to search for, read [how to search the text in your photos](https://firelet.media/blog/search-text-in-your-photos).

## Questions

### Which languages can it read?

On Windows it uses the text recognition built into Windows, in the first of your Windows languages that has it installed, falling back to English. The macOS and Linux builds, when available, read English.

### Is my text sent to a server?

No. Text is recognised on your PC and stored in your library; nothing is downloaded or sent.

### Do I have to run Read text again for new photos?

Yes. Click Read text after adding pictures; it reads only the photographs it hasn't read yet.