public final class TextScript

Writing system selectors for TextRecognizer. The recognizers behind the portable API are script-specific rather than language-specific: one script covers every language written in it, and each script model also recognizes Latin characters embedded in the page.

Each selector call is also a build-time dependency marker, exactly like VisionBackends. The methods must remain distinct so a build packages only the script models the application actually asks for; a Japanese OCR application does not carry the Korean or Devanagari model.

TextRecognizer recognizer = new TextRecognizer(
        new VisionOptions().textScript(TextScript.japanese()));
recognizer.process(VisionImage.encoded(jpeg)).ready(result -> {
    Log.p(result.getText());
});

A selector states the script, not the backend. Android maps each one to the matching ML Kit recognizer. Apple platforms map it to the Vision recognition languages for that script, and report the analysis as unsupported when the OS does not recognize that script – selecting VisionBackends.mlKitTextRecognition() then supplies the script model through ML Kit instead.