Summarize Text from Image Privacy

Get the text out of images, screenshots, and scanned PDFs

Snip a region of the page, click an image, or drop in a file. The text appears in the side panel, ready to edit, copy, or save.

Coming to the Chrome Web Store

The side panel open beside a webpage A browser window with a dashed box drawn around part of a picture, and the extension side panel docked on the right showing the recognised text as lines of prose, with a Copy button under them. Capture text Raw Tidy Copy

Features

Snip a region

Drag a box around anything you can see on the page: part of a chart, a line in a video frame, one row of a table. Only what is inside the box gets read.

Click an image on the page

Images light up as you move the pointer over them. Click the one you want and the panel starts reading it.

Right-click any image

Right-click an image and choose "Extract text from image". The side panel opens and the text arrives there.

Capture the visible page

Reads whatever is on screen right now, including the page you are looking at in Chrome's built-in PDF viewer.

Files and paste

Drop an image or a PDF onto the panel, open one from your computer, or press Ctrl+V to paste an image you already copied.

Scanned PDFs

Read all pages or a range like 1-3,5. Up to 50 pages per run and files up to 100 MB. Password-protected files ask for the password, and you can cancel a long run and start it again. A wide range is rendered at a lower resolution, and the panel says so when that happens.

Latin and Cyrillic

Reads both alphabets, including both in the same image. Recognition runs on your device.

Keeps the shape of the text

Paragraphs stay paragraphs, list items stay on their own lines, and columns in a table-like block stay apart instead of running together.

Raw and Tidy

Raw is exactly what the recognizer produced. Tidy removes stray spacing and broken line endings without changing the words. One switch moves between them, and a setting lets Tidy merge wrapped lines into full paragraphs.

Edit before you copy

The text in the panel is editable. Fix a misread character or delete a line you do not need, then copy the corrected version.

Copy, automatically if you want

One button puts the current text on the clipboard. Turn on "Copy automatically" and every finished result lands there by itself.

Export to a file

Save the text as .txt or .md. When the result is laid out like a table, a .csv export appears as well, so the columns open as columns in a spreadsheet.

History of the last 50 results

Each result is kept in this browser profile with a small preview and its text. Search them, pin the ones worth keeping, reopen or delete one, or clear the whole list.

Summarize

Sends the extracted text to an online service and brings back a short summary. It runs only when you press the button.

Translate into 23 languages

Arabic, Chinese, Czech, Danish, Dutch, English, Finnish, French, German, Greek, Hindi, Hungarian, Italian, Japanese, Korean, Polish, Portuguese, Romanian, Russian, Spanish, Swedish, Turkish, Ukrainian. You pick the language, and long text is sent in pieces, so if one piece fails, Retry continues from there instead of starting over.

Keyboard shortcuts

Alt+Shift+I opens the panel on any page. Alt+Shift+S starts a region snip.

How it works

  1. Open the side panel from the toolbar icon, or with Alt+Shift+I.
  2. Choose what to read: snip a region, click an image, right-click an image, capture the visible page, or drop in an image or PDF.
  3. The text appears in the panel, with the source name and how confident the read was.
  4. Switch between Raw and Tidy, and fix anything that came out wrong.
  5. Copy the text, export it as a file, or send it to Summarize or Translate.

Questions

What kind of text can it read?

Text written in Latin or Cyrillic characters, in images, screenshots, page regions, and scanned PDFs. A larger, sharper source reads better than a small or blurry one, so a tighter snip around the text usually gives a better result than a full-page capture.

Do I have to upload anything?

No. You point at what is on your screen, or open a file from your computer. Summarize and Translate are the only parts that use an online service, and they send the extracted words, not the picture.

Are Summarize and Translate available right now?

They are built into the extension, but the online service behind them is not switched on yet, so those two buttons are not usable at the moment. Everything else, including reading, editing, copying, exporting, and history, works without any service.

Why does it ask for page access?

Nothing is requested when you install it. Access is asked for once, when you first snip or capture on a website, because the extension has to see the page to read the part you selected.

How many pages of a PDF can it handle?

Up to 50 pages in one run, from a file of up to 100 MB. Pages are processed one at a time, you can cancel partway through, and you can pick a range instead of the whole document.

Can I fix a mistake in the recognized text?

Yes. The result is an editable field. Your corrections are what gets copied, exported, and kept in history.

What can I save the text as?

.txt, .md, and, when the result looks like a table, .csv. The .csv opens correctly in Excel, including text with accented or Cyrillic characters.

Does the history fill up my disk?

It keeps up to 50 results and up to 25 MB, storing a small preview image rather than the full screenshot. Older items drop off on their own, pinned ones stay, and you can clear everything in one action.

Which browser do I need?

Chrome 140 or newer.