OCR languages
Choose recognition languages, set defaults, and download on-demand language packs.
Telling Scanix which languages a page contains makes recognition much faster and more accurate. This reference covers the Default OCR languages picker, the on-demand language packs, and the per-template language override. For the bigger picture of how OCR fits together, see the OCR overview.
Where languages are set
| Level | Location | Wins over | Use when |
|---|---|---|---|
| App default | Settings → OCR Engine → Default OCR languages | The fallback when nothing more specific is set | You want one usual set of languages for most documents |
| Per template | Template Designer (captured with the template) | The app default | A document type always uses the same language(s) |
| Per OCR run | The Select OCR languages prompt in the viewer | A one-off choice for this run | An untemplated import whose language differs from your default |
The more specific level wins: a template's language overrides the app default, and a per-run choice overrides both. Most operators set the app default once and rarely touch the rest. The full Settings reference is in OCR engine & languages settings.
Set your default OCR languages
The Default OCR languages are used whenever an OCR run doesn't specify its own. Setting them is a staged change, so it only takes effect after you click Apply.
Open the Default OCR languages picker
Open Settings, click the OCR Engine tab, and scroll down to the Default OCR languages section (directly under the engine cards).
Pick your languages
Click any chip in the Popular group, or type in the Search languages… box and click chips in the All languages list. You can select as many as you need — the All languages column shows a running N selected count.
Apply
Click the global Apply button in the Settings footer to save your choice. Any selected languages whose packs aren't yet on disk are queued for download at this point.
Screenshot
Settings open on the OCR Engine tab, scrolled to Default OCR languages — the Popular chips on the left with two selected (check marks), and the All languages search list on the right showing a N selected count. — shot ocr-languages-01
Picker controls
| Control | What it does |
|---|---|
| Popular chips | Quick picks: English, French, German, Spanish, Arabic, Russian, Chinese, Japanese, Korean, Italian |
| All languages list | The full searchable list of selectable languages |
| Search languages… | Filters the All languages list as you type |
| Language chip | Click to select (shows a check mark); click again to deselect |
| N selected | Running count of chosen languages in the All languages column |
| Status line | Green dot + "Scanning will be much faster." when at least one language is chosen; amber dot + "Scanning will be slower on every page." when the list is empty |
| Advanced: use auto-detect instead | Clears all selections (confirm with Yes, clear) to switch to auto-detect |
| Apply | Saves the staged selection and queues any missing packs for download |
Empty list means slower auto-detect
| Default OCR languages | What happens | Speed |
|---|---|---|
| One or more languages selected | Scanix loads exactly those models | Faster, more accurate |
| Empty | Scanix falls back to auto-detect and runs an extra wide-net script-detection pass to guess the script first | Slower on every page |
Leaving the list empty is valid, but it makes every page slower. Whenever you know the languages your documents use, select them. The picker's status line tells you which state you're in at a glance.
Empty = slower
With no default languages set, Scanix has to detect the script on every page before it can read it. Choosing even one language removes that extra pass.
Download on-demand language packs
Many common languages are bundled and work fully offline. Others download the first time you use them as small language packs. A non-bundled language chip carries a small download glyph and the tooltip "On-demand pack — downloads on first use."
You don't have to download anything manually before selecting a language — picking it and clicking Apply queues the pack, with a disk-space pre-check that warns you if there isn't room. Already-installed languages confirm with a toast such as "<Language> already installed — ready to use immediately."
Pack states on a language chip
| State | Appearance | What it means |
|---|---|---|
| Bundled | Plain chip; check mark when selected | Installed and ready offline |
| On-demand (not installed) | Download glyph; tooltip "On-demand pack — downloads on first use" | Will download when you Apply with it selected |
| Downloading | Chip fills like a progress bar with bytes/percent (for example 12.3 / 40 MB); hover shows a ✕ to cancel | The pack is being fetched |
| Failed | Chip turns red; tooltip "Download failed — click to retry" | Click the chip to retry |
| Just finished | Brief check, "<Language> ready" | The pack is now installed |
Some engines flag a pack download too
On the OCR Engine cards above, a "Needs download" badge means the engine's runtime is present but its language weights aren't on disk yet. Select a language pack listed under that engine and use the Download action to fetch it, then confirm the Engine Status panel reads Ready. See OCR engine & languages settings.
What happens if a language pack is missing at run time
If a requested language's pack isn't available when OCR runs, Scanix continues with the languages it does have rather than stopping — but accuracy for the missing language may drop. Apply your language choice ahead of a big batch so the packs are in place first.
Per-template language override
A template can carry its own language so every document captured or imported with it is read the same way, without relying on the app default. When a template has a language set, Scanix uses it automatically and does not show the per-run Select OCR languages prompt for those imports.
| Setting | Where | Default | Effect |
|---|---|---|---|
| Template language | Captured with the template in the Template Designer | Inherits the app default | Pins the OCR language(s) for that template's documents |
| OCR Engine (template default) | Template Designer → OCR Behaviour card | App default (Settings → OCR Engine) | Optionally pins the engine too, alongside the language |
The engine dropdown's options are App default (Settings → OCR Engine), Tesseract, EasyOCR, PaddleOCR, and Scanix Vision (in-house). Leave it on App default to inherit your global choice. (Availability of PaddleOCR and Scanix Vision (in-house) depends on the online catalog — Tesseract and EasyOCR are always available.)
The per-run language prompt
For untemplated imports with no language bound, Scanix asks which languages to use right before OCR starts, so it loads the correct models instead of guessing.
| Element | Detail |
|---|---|
| When it appears | You click OCR in the viewer on a free-form import that has no template language set, and you haven't already remembered a choice this session |
| Title | Select OCR languages |
| Picker | Multi-select Languages list (with RTL/CJK badges and a warning if a pack isn't installed) |
| Remember for this session — reuse for the next OCR run without asking again | Checkbox, ticked by default; cleared when you leave the viewer |
| Buttons | Cancel and Run OCR (Run OCR stays disabled until you pick at least one language) |
Template imports skip this prompt entirely and use the template's pinned language.
Screenshot
The Select OCR languages modal over the viewer — a multi-select language list with two languages chosen, the Remember for this session checkbox ticked, and the Run OCR button enabled. — shot ocr-languages-02
Next steps
OCR overview
How OCR works in Scanix, the engines, and the searchable text layer.
OCR engine & languages settings
The full Settings reference for the OCR Engine tab, engines, and language packs.