What does the OCR language setting do, and can I change it?

Last reviewed: October 29, 2026 · Markdown version

Text recognition does not read letters in a vacuum. It matches shapes against a model trained on one writing system, and many engines then check the result against the words of a language. The language setting chooses which model and which word list. Adobe's Acrobat help puts it in one line: "Document Language - Sets the OCR language for accurate character detection" (Scanned PDF settings, Adobe Acrobat Help, read 29 October 2026). Whether you can change it depends on the app, and in some apps there is nothing to change.

What a wrong setting does

Adobe's phone scanner documents the worst case. "If you’ve selected a language to recognize that doesn’t match the document language, OCR produces no output. For example, the scanned document is German and you set Danish as the text recognition language." The same page lists "A source document with multiple languages" among the conditions that "may impair OCR quality" (Troubleshoot — Adobe Scan for iOS, Adobe, read 29 October 2026). So a page that comes back empty is not always a bad scan; check the language first. For how good a correct setting gets you, see how accurate OCR is.

Where to change it

Tool	Is there a setting?	Where
Adobe Acrobat (computer)	Yes, each time you run OCR	Scan & OCR dialog; also under Settings
Adobe Scan (phone)	Yes, one default	Preferences, Text recognition language
Google Drive	No; it detects the language	Nothing to set
Apps built on Apple Vision or Google ML Kit	Decided by the app's developer	Only if the app exposes it

Acrobat. When you recognise text in one file, "In the dialog box, select the page range and language for text recognition" (Recognize text in scanned PDFs with Acrobat, Adobe Acrobat Help, read 29 October 2026). When scanning straight into Acrobat, Adobe says to "Use Settings to choose language and output type" (the Scanned PDF settings page). Neither page lists the languages; the steps themselves are on OCR a PDF in Adobe.

Adobe Scan. "The default text recognition language is set to the language of your device." To change it, open Preferences and "tap Text recognition language and tap a desired language" (Configure Adobe Scan — Adobe Scan for Android, Adobe, read 29 October 2026; the iPhone page says the same). Its list has 18 languages and no Indian one; more on that app in whether Adobe Scan does OCR.

Google Drive. There is no switch to find. Google's tips for converting a scan say "Google Drive detects the language of the document", and link a list of supported languages that includes Hindi (Convert PDF and photo files to text, Google Drive Help, read 29 October 2026). The same tips ask for text "at least 10 pixels high" and a file of "2 MB or smaller".

Inside the engines phone apps use

On iPhone and iPad, apps that read text with Apple's Vision framework pass it a list: "An array of languages to detect, in priority order", where "The order of the languages in the array defines the order in which languages are used during language processing and text recognition" (recognitionLanguages, Apple Developer Documentation, read 29 October 2026). An app can instead ask Vision to work it out, a switch Apple describes as whether "to attempt detecting the language to use the appropriate model for recognition and language correction" (automaticallyDetectsLanguage, Apple Developer Documentation, read 29 October 2026).

Google's ML Kit, common on Android, splits by writing system rather than by language. Its developer guide lists a separate library for each: one comment reads "To recognize Latin script", others name Chinese, Devanagari, Japanese and Korean (Recognize text in images with ML Kit on Android, Google for Developers, read 29 October 2026). Within a script, Google sorts languages into three levels, and the third is a reminder that a language name in a menu is not a promise: "Mapped languages are those supported by mapping them to another language code or to a general character recognizer", so "en-GB" is "not treated any differently than "en"" (Text recognition v2: supported languages, ML Kit, Google for Developers, read 29 October 2026). Our reading: the scripts an Android app built on ML Kit can read are chosen by its developer when the app is built.

"Reference language" and language packs

Both phrases turn up in searches. None of the Adobe, Google or Apple pages cited above uses the term "reference language", and none describes installing an OCR language pack (all read 29 October 2026). If an app offers either, its own help is the only reliable description of what it does.

DocFind has no language setting

There is nothing to change in DocFind, and here is what that means in practice. On iPhone and iPad it passes Apple's Vision the whole set of languages Vision reads, with automatic detection on; that set has 30 languages and no Hindi or any Devanagari. On Android it uses Google's ML Kit model bundled into the app, which reads Latin-script printed text only. So scanned Chinese, Arabic, Japanese, Korean or Thai becomes searchable on iPhone and iPad but not on Android, and scanned Hindi or any Devanagari does not become searchable on either. Can OCR read a Hindi PDF? compares other tools for that case.