Languages

Languages

Bundled (Free)

The Free AAR ships English (eng.traineddata, from tessdata_fast, ~4 MB) inside assets/tessdata/. It is extracted to app-private storage on first initialize() — nothing to download, works offline.

ocr.initialize(context, language = "eng")

Adding a language at runtime

Download any *.traineddata (from tessdata_fast, tessdata, or tessdata_best), then register it:

// e.g. after downloading fra.traineddata to a File
ocr.addLanguage(context, language = "fra", file = downloadedFile)
ocr.initialize(context, language = "fra")

addLanguage copies the file into the engine's tessdata dir. On the next initialize, that code becomes usable.

Multiple languages at once

Join codes with +:

ocr.initialize(context, language = "eng+fra")

Tesseract will recognise text in any of the listed languages. More languages = slower init and slightly slower recognition.

Which model quality?

Pack	Size / lang	Speed	Accuracy
tessdata_fast	~1–5 MB	Fastest	Good (Free default)
tessdata	~10–15 MB	Medium	Better
tessdata_best	~15–30 MB	Slowest	Best (Pro default)

Full language list & downloads: https://github.com/tesseract-ocr/tessdata_fast

Checking what's installed

val langs = ocr.getAvailableLanguages(context)   // e.g. ["eng", "fra"]

Pro

Tesseract Pro bundles 12 languages using tessdata_best (highest accuracy) — no runtime download needed.