Text recognition

Recognise on-screen text with Apple's Vision framework, entirely on the device — no network and no screenshot upload. Prefer element lookups (clickElement) where they work: the element tree is structured, fast and exact. OCR is for text drawn into images or canvases that the element tree does not contain.

Aperçu

·Set the language and speed once with ocrConfigure; the whole task uses them
·Chinese needs iOS 14 or later, Japanese / Korean / Russian / Thai / Vietnamese need iOS 16 or later, and Vision does not support Arabic
·Coordinates come back in points, ready for client.click

ocr#

Recognise the text on screen and return it line by line in reading order.

typescript

ocr(region?: [number, number, number, number] | null): Line[] & { text: string };

type Line = {
  text: string;
  confidence: number;
  x: number; y: number; width: number; height: number;
  centerX: number; centerY: number;
};

Paramètres

Nom	Type	Obligatoire	Par défaut	Description
`region`	`[x, y, w, h] \| null`	Non	—	Only read this area; omit it for the whole screen. A smaller area is several times faster. Written as in colour and image matching: 0–1 screen ratios when all four numbers are between 0 and 1, otherwise points.

Valeur de retour

Type	Description
`{ text, confidence, x, y, width, height, centerX, centerY }[]`	One item per line, coordinates in points. The array also carries .text: every line joined with newlines.

Exceptions

Exception	Levée quand
`WDAError`INVALID_ARGUMENT	An argument is wrong: an empty keyword, a wait that is not a number or exceeds 600, a region that is not [x, y, w, h], an old-style {}, and so on.
`WDAError`OCR_LANG_UNSUPPORTED	This device's Vision framework does not support the configured language.
`WDAError`FEATURE_UNAVAILABLE	The app on the device is too old to support text recognition.

Exemple

javascript

var lines = ocr([0.05, 0.12, 0.9, 0.25]);
logi(lines.text);
for (var i = 0; i < lines.length; i++) {
  logi(lines[i].text + ' @' + lines[i].centerX + ',' + lines[i].centerY);
}

Voir aussi findText、ocrConfigure

findText#

Find a piece of text on screen and return the line that contains it.

typescript

findText(keyword: string | RegExp, timeoutSec?: number, region?: [number, number, number, number] | null, exact?: boolean): Line | null;

Paramètres

Nom	Type	Obligatoire	Par défaut	Description
`keyword`	`string \| RegExp`	Oui	—	The text to find (a line containing it counts), or a regular expression. Must not be an empty string.
`timeoutSec`	`number`	Non	`0`	The longest wait, in seconds. 0 means look once; above 0 it re-captures and re-reads every 0.5 seconds until found or out of time.
`region`	`[x, y, w, h] \| null`	Non	—	Only read this area; omit it for the whole screen. A smaller area is several times faster. Written as in colour and image matching: 0–1 screen ratios when all four numbers are between 0 and 1, otherwise points.
`exact`	`boolean`	Non	`false`	Pass true to require the whole line to match, not just contain the keyword.

Valeur de retour

Type	Description
`{ text, confidence, x, y, width, height, centerX, centerY } \| null`	The matching line, or null.

Exceptions

Exception	Levée quand
`WDAError`INVALID_ARGUMENT	An argument is wrong: an empty keyword, a wait that is not a number or exceeds 600, a region that is not [x, y, w, h], an old-style {}, and so on.
`WDAError`OCR_LANG_UNSUPPORTED	This device's Vision framework does not support the configured language.
`WDAError`FEATURE_UNAVAILABLE	The app on the device is too old to support text recognition.

Exemple

javascript

var hit = findText('Balance', 5);
if (hit) {
  logi(hit.text);
}
var price = findText(/\$\s*[\d.]+/);

Voir aussi clickText、ocr

clickText#

Wait for a piece of text to appear and tap it, at a slightly random spot within the line.

typescript

clickText(keyword: string | RegExp, timeoutSec?: number, region?: [number, number, number, number] | null, exact?: boolean): boolean;

Paramètres

Nom	Type	Obligatoire	Par défaut	Description
`keyword`	`string \| RegExp`	Oui	—	The text to find (a line containing it counts), or a regular expression. Must not be an empty string.
`timeoutSec`	`number`	Non	`10`	The longest wait, in seconds. It re-captures and re-reads every 0.5 seconds.
`region`	`[x, y, w, h] \| null`	Non	—	Only read this area; omit it for the whole screen. A smaller area is several times faster. Written as in colour and image matching: 0–1 screen ratios when all four numbers are between 0 and 1, otherwise points.
`exact`	`boolean`	Non	`false`	Pass true to require the whole line to match, not just contain the keyword.

Valeur de retour

Type	Description
`boolean`	True when tapped, false when the text never appeared.

Exceptions

Exception	Levée quand
`WDAError`INVALID_ARGUMENT	An argument is wrong: an empty keyword, a wait that is not a number or exceeds 600, a region that is not [x, y, w, h], an old-style {}, and so on.
`WDAError`OCR_LANG_UNSUPPORTED	This device's Vision framework does not support the configured language.
`WDAError`FEATURE_UNAVAILABLE	The app on the device is too old to support text recognition.

Remarques

By default the task log records whether the text was tapped or not found; turn it off with setClickLog(false).

Exemple

javascript

clickText('Claim now', 8);

Voir aussi findText、clickElement

ocrConfigure#

Set this task's recognition languages and speed, validating the languages up front. Optional — the default is Simplified Chinese plus English.

typescript

ocrConfigure(languages?: string[] | string, level?: "accurate" | "fast", correction?: boolean): { ok: true; languages: string[]; level: string };

Paramètres

Nom	Type	Obligatoire	Par défaut	Description
`languages`	`string[] \| string`	Non	—	The recognition languages, for example ['zh-Hans', 'en-US'] or ['ja-JP', 'en-US']; a single language can be a plain string.
`level`	`"accurate" \| "fast"`	Non	`"accurate"`	accurate is precise, fast is quick.
`correction`	`boolean`	Non	`false`	Pass true to let the system correct results against its dictionary.

Valeur de retour

Type	Description
`{ ok, languages, level }`	The settings now in effect.

Exceptions

Exception	Levée quand
`WDAError`OCR_LANG_UNSUPPORTED	This device's Vision framework does not support the configured language.
`WDAError`INVALID_ARGUMENT	The languages are not an array of strings, level is not accurate / fast, or an old-style {} was passed.
`WDAError`FEATURE_UNAVAILABLE	The app on the device is too old to support text recognition.

Remarques

Its value is failing early: without it, a device that cannot read Chinese only shows the problem deep inside a loop.

Exemple

javascript

ocrConfigure(['zh-Hans', 'en-US'], 'accurate');

Voir aussi ocrCapabilities、ocr

ocrCapabilities#

Which languages this device can actually recognise.

typescript

ocrCapabilities(): { engine: string; revision: number; accurate: string[]; fast: string[]; supportsChinese: boolean };

Valeur de retour

Type	Description
`{ engine, revision, accurate, fast, supportsChinese }`	accurate / fast list the language codes each mode supports.

Exceptions

Exception	Levée quand
`WDAError`FEATURE_UNAVAILABLE	The app on the device is too old to support text recognition.

Exemple

javascript

logi(JSON.stringify(ocrCapabilities()));

Voir aussi ocrConfigure