Quick Start

Quick Start

1. Add the dependency

// build.gradle.kts (app)
dependencies {
    implementation("dev.ffmpegkit-maintained:tesseract-android:5.5.0")
}

No mavenCentral() tweaks needed — it's a normal Central artifact. For JitPack instead:

// settings.gradle.kts
repositories { maven("https://jitpack.io") }
// build.gradle.kts (app)
implementation("com.github.ffmpegkit-maintained:TesseractOCR:5.5.0")

2. Run OCR

import dev.ffmpegkit.tesseract.TesseractOCR

class MyViewModel : ViewModel() {
    private val ocr = TesseractOCR()

    suspend fun read(bitmap: Bitmap): String {
        if (!ocr.isInitialized) {
            ocr.initialize(context, language = "eng")   // English is bundled
        }
        return ocr.recognize(bitmap).text
    }

    override fun onCleared() { ocr.release() }
}

initialize, recognize, and recognizeFile are suspend functions — call them from a coroutine. They already move work to a background dispatcher, so you don't need withContext yourself.

3. Clean up

Call ocr.release() when done (e.g. onCleared() / onDestroy()) to free native memory. A released instance can be re-initialised.

Notes

• A single TesseractOCR instance is not thread-safe — serialise calls or use one instance per worker.
• The bundled model is tessdata_fast English. For other languages see Languages.
• See Performance for choosing the right Page Segmentation Mode.