docAnalyzer

Enhanced OCR

Premium per-page OCR for documents where the automatic pass wasn't good enough.

Enhanced OCR is the higher-accuracy pass you trigger manually when Automatic OCR couldn't quite read a document (handwriting, low-quality scans, dense or complex layouts). It's credit-metered: 1 credit per page.

When to use it

Try Enhanced OCR when:

  • The document already exists in your workspace.
  • Automatic OCR ran but produced garbled, incomplete, or missing text.
  • The pages contain handwriting, dense tables, multi-column layouts, marginalia, or unusual fonts.
  • Citations to the document return nothing or quote nonsense, usually a sign the underlying OCR is poor.

If automatic OCR worked fine, you don't need Enhanced. The two produce text the same way downstream; Enhanced just produces better-quality text on harder inputs.

Run Enhanced OCR

  1. 1

    Open the document from the documents page or the info panel.

  2. 2

    Open the action menu (the kebab menu on the row, or the menu in the info panel).

  3. 3

    Click Run Enhanced OCR.

  4. 4

    Confirm. The cost is shown up-front (1 credit × page count).

  5. 5

    Wait for the re-indexing pass to finish.

You can trigger it on multiple documents at once: multi-select rows on the documents page, then Bulk actions → Enhanced OCR.

What happens

Enhanced OCR runs a higher-accuracy OCR pass and re-indexes the document text. Citations from new chats land against the refreshed text. Older chats that referenced the document still work (citations still resolve) but they reference the previous OCR state.

The original scan stays in the viewer. Enhanced OCR replaces the extracted text, not the displayed pages.

Cost: 1 credit per page

Every page in the document is billed, even pages where the automatic pass was already good and enhanced OCR doesn't change much. The cost is per page in the document, not per page Enhanced OCR "improves".

If you're not sure whether Enhanced will help, try a single short document first; compare a Focus answer against the original document; scale up if Enhanced is clearly producing better citations.

Languages

Enhanced OCR supports the same 40+ languages as the automatic pass, with better accuracy on handwriting and complex scripts. It's still best-effort: there's no OCR system that handles every document perfectly.

What it isn't

  • Not a guarantee. Pages where the scan is genuinely too poor still won't extract well. Enhanced is better, not magic.
  • Not free. Every page costs 1 credit, no refund if quality didn't improve.
  • Not automatic. Has to be explicitly triggered; there's no "always use Enhanced OCR" setting. You can run it on many documents at once by multi-selecting on the documents page and launching from bulk actions, but it never fires on its own.

What's next

Was this helpful?