Extract Text

Get the text out of images — receipts, screenshots, photographed documents, scanned pages, anything. Outputs .txt (plain) or .docx (formatted paragraphs).

This page does not translate — it just extracts. Pipe the output into Translate Document if you also want translation.

Two extraction methods

Method	Best for
OCR	High-volume / batch / cost-sensitive (free or near-free per image)
LLM vision	Layout preservation, mixed scripts, low-quality images, handwriting

Pick the default in Settings → Extract Text → Extraction method.

OCR engines (OCR method)

Engine	Cost	Offline	Languages	Notes
Tesseract	Free	Yes	100+	Default. Needs a system install.
EasyOCR	Free	Yes (after model download)	80+	Best for non-Latin scripts. ~1 GB models.
Google Cloud Vision	Paid (1,000 free / month)	No	60+	Highest accuracy.

Configure in Settings → OCR.

Walkthrough

1. Click Extract Text in the sidebar.
2. Drop one or more image files (.png, .jpg, .jpeg, .bmp, .webp, .tiff, .tif).
3. Pick the Source language (helps OCR pick the right model).
4. Pick the Output format — .txt or .docx.
5. Click Extract (or Ctrl+Enter).
6. Open the row when done.

When to use which

• Text-heavy receipt / invoice → Tesseract is fast and accurate.
• Photographed handwritten notes → LLM vision wins by a lot.
• Manga / comic panels → EasyOCR (handles vertical CJK text well).
• Form with lots of small fields → Google Cloud Vision tends to preserve field boundaries better than the others.

Tips

OCR or LLM, not both

The page picks one method and runs it. To compare outputs, run the same image twice with different methods.

Setup-required dialog

If you pick OCR but no OCR engine is configured (or LLM but no LLM key is configured), the page surfaces a single "Setup Required" dialog that links straight to the relevant Settings tab.

Shortcuts

Shortcut	Action
Ctrl+Enter	Extract
Ctrl+O	Browse
Ctrl+F	Focus history search