OCR in Scanix
How Scanix recognizes text — the Tesseract and EasyOCR engines and the searchable text layer.
A scan is just a picture of a page — your computer can't search it, select its words, or copy a phrase out of it. OCR (optical character recognition) is how Scanix Desktop reads that picture and turns it into real, machine-readable text. This page explains what OCR produces, the engines that do the reading, and where each choice lives, then points you to the how-to pages for the day-to-day tasks.
What OCR actually produces
When Scanix reads a page, it doesn't replace your scan with text — it builds a searchable text layer and lays it invisibly over the image. The page still looks exactly like the original scan, but underneath every printed word there's now a real, selectable character. That single idea is the foundation for almost everything downstream:
- You can search inside the document and jump to matches.
- You can select and copy text straight off the page, including non-Latin scripts such as Arabic.
- Your exported file becomes a Searchable PDF — it looks identical to the scan, but anyone can find and copy text in it with an ordinary PDF reader.
- Capture zones can pull index values (an invoice number, a date) out of the recognized text instead of making someone type them.
Because the text sits in a hidden layer rather than redrawing the page, recognition can be wrong without changing how the document looks — which is exactly why Scanix lets you read, search, and correct that layer region by region after the fact. See Edit & correct text.
OCR reads text — it doesn't classify or extract on its own
Plain OCR produces the text layer and nothing more. Pulling specific fields out of a document, or deciding what kind of document it is, is the job of templates and AI capture, which build on top of the recognized text.
The engines: Tesseract and EasyOCR
The actual reading is done by an OCR engine — the recognition technology Scanix runs against each page image. Two engines are always bundled and work fully offline, with no setup and no internet:
| Engine | Best for | Notes |
|---|---|---|
| Tesseract | Invoices, receipts, contracts, books | Fast and reliable for clean printed text. The default. |
| EasyOCR | Receipts, mail, mixed scans | Handles rotated pages, mixed fonts, and lower-quality scans. |
Think of it as a trade-off: Tesseract is the quick, dependable choice for tidy printed pages, while EasyOCR is the more forgiving choice when scans are crooked, mixed, or a bit rough. Both ship inside Scanix, so you can switch between them at any time without downloading anything.
Other engines appear when the platform publishes them
You may sometimes see additional engines such as PaddleOCR or Scanix Vision (in-house) offered — for example in a template's engine dropdown. These are delivered through Scanix's online catalog rather than bundled in the installer, so they are only usable once the platform has published them and (where needed) you've downloaded their language data. Tesseract and EasyOCR are the two you can always count on.
A card may carry a status badge — Needs download (the engine is present but its language weights aren't on disk yet), Coming soon (the engine isn't available to run), or a "requires a newer Scanix" note. An engine can be selected but still need a language pack before it can actually read a page; the Engine Status panel under the cards tells you when it shows Ready.
Where the engine is chosen
There are two levels at which you decide which engine reads your pages, and the lower level wins.
App default — Settings
Sets the engine used whenever nothing more specific is chosen. Open Settings, then the OCR Engine tab, and click the engine card you want.
Per template
A template can pin its own engine so every document of that type is read consistently, overriding the app default.
The app-wide default lives at Settings → OCR Engine. The section is headed OCR Engine with the sub-label "Select the OCR engine used for text recognition," and the engines appear as selectable cards. Click a card and the choice is saved immediately; Tesseract is the default. This is the engine Scanix uses any time a more specific setting hasn't overridden it.
A template can pin its own engine in the Template Designer, on the OCR Behaviour card, via the OCR Engine (template default) dropdown. Leaving it on App default (Settings → OCR Engine) inherits the global choice; picking a specific engine forces it for every document captured or imported with that template. (Individual capture zones can override the engine once more from their own settings — useful when one field on a page reads better with a different engine.)
The mental model is a simple cascade: a zone's setting beats the template's, the template's beats the app default. Most operators set the app default once and never touch the rest.
Screenshot
Settings open on the OCR Engine tab, showing the Tesseract and EasyOCR engine cards (Tesseract selected), with the Engine Status panel below reading Ready. — shot ocr-overview-01
How a page gets read
You don't have to babysit OCR. Recognition can happen in two ways, and which one fires depends on how the document arrived:
- Automatically, right after import or scan — but only when the template you used opts in. A template's Auto-run OCR after import/scan toggle turns this on; Scanix then holds OCR until the pages have actually rendered on screen (you'll see a brief Preparing pages… state) so the window never freezes with rendering and reading fighting for the processor. Documents opened without a template never auto-read.
- On demand, by clicking the OCR button in the document viewer. While a page is being read, the OCR button itself fills like a progress bar, so you can watch it work without losing the button.
Either way, the result is the same searchable text layer written onto the document's pages — nothing is exported to disk until you Process or export. The details of running, re-running, and skipping recognition are in Run OCR.
Languages matter as much as the engine
An engine can only read scripts it has language data for, and telling Scanix which languages a page contains makes recognition much faster and more accurate. Set your usual languages once under Default OCR languages (on the same OCR Engine tab), and Scanix uses them whenever a run doesn't specify its own. Leave the list empty and Scanix falls back to a slower auto-detect pass that has to guess the script first.
Many common languages are bundled and work offline; others download on first use as small language packs. For the full picture — picking languages, the on-demand packs, and the per-run prompt you'll see for untemplated imports — see OCR languages.
The text layer is yours to correct
OCR is rarely perfect on difficult scans, so the recognized text is fully editable. After a page is read, every region appears in the viewer's Text tab, where you can search it, step through low-confidence regions flagged for review, fix the wording, retype a region, or adjust its box on the page. By default these corrections save instantly; if you prefer a deliberate commit step, you can switch on an explicit Save button instead.
Correcting the text doesn't just tidy the on-screen list — it rebuilds the hidden searchable layer that ends up in your exported PDF, so a fix you make here is a fix anyone searching the final file benefits from. The full workflow, including auto-save versus manual save, is in Edit & correct text.
From recognized text to a searchable file
When you process a document, Scanix's default output is a Searchable PDF: a file that looks identical to your scan but carries the invisible, selectable text layer built from OCR. That's what makes the exported document findable and copyable long after it leaves Scanix.
For large archives, an optional high-compression (MRC) mode can make those PDFs dramatically smaller by keeping crisp text on a downsampled colour background. It's off until you enable it, and then offers a range of quality levels. How to turn it on and which level to pick is covered in Searchable PDF & compression.
Frequently asked
Next steps
OCR languages
Set default languages, add on-demand packs, and understand the per-run prompt.
Run OCR
Read pages on demand or automatically after import, and re-run when needed.
Edit & correct text
Review, search, and fix the recognized text layer region by region.
Searchable PDF & compression
Export a selectable, searchable PDF and shrink it with high-compression mode.