Extract native PDF text and every word inside mixed images

Native text is read directly, while every eligible page image is recognized with PP-OCRv6. Overlap with hidden OCR layers is deduplicated.

Native text + OCRMixed text and imagesReading-order reconstructionHTML / TXT / Markdown

Open PDF content extraction →Read how it works

Capability quick facts

These facts describe the current product, not unshipped roadmap work.

Text sources: Native PDF text plus image OCR
Models: Auto / Fast / Professional / Ultimate
Mixed layout: Merges text and images by page coordinates
Outputs: Long HTML / TXT / Markdown

Reviewed by the DataDance product and engineering team · Published 2026-08-29 · Updated 2026-08-30

Finish in three steps

Confirm important settings before export; the processing location is always disclosed.

Choose a PDF

Read text coordinates, image positions and shared resources per page.

Choose an OCR tier

Auto uses Professional on desktop and lower-memory Fast on mobile; desktop can also select Fast, Professional or Ultimate.

Review and export

Merge native text, images and non-duplicate OCR into a self-contained long page.

Does it work with text PDFs and scans?

Yes. Native words come from PDF text operators; scans and images inside mixed pages use ImgIng PP-OCRv6. Both streams are merged by page coordinates.

How are OCR models selected and loaded?

Auto chooses Professional on desktop and Fast on mobile; desktop users can choose Fast, Professional or Ultimate. On first use the selected model is downloaded and verified from multiple model CDN sources, automatically falls back when one fails, and is then reused from the browser cache.

What if a scan already has an invisible OCR layer?

Normalized OCR output is compared with the page’s native text. Existing phrases are not inserted twice, while unmatched text and the source image remain available for review.

Why export long-page HTML?

HTML keeps page markers, paragraphs, images and recognized image text together. TXT and Markdown provide lighter content-only alternatives.

Updated 2026-08-30 · Live capability detection inside the tool is authoritative

Frequently asked questions

These visible answers match the current product behaviour and structured data.

Is text inside images recognized?

Yes. Every eligible page image is OCR-processed and merged with native text.

Will hidden OCR text duplicate?

Overlapping normalized text is removed automatically.

Does processing upload the PDF?

No. PDF parsing, OCR, layout reconstruction and export stay local.

Explore ImgIng

Each task page documents real settings, limits and format advice—not keyword-swapped duplicates.