Home

Awesome

google-translate-api-x

Actions Status NPM version

A free and unlimited API for Google Translate :dollar: :no_entry_sign: written with compatibility in mind, made to be crossplatform. Now with over 100 new languages supported!

Features

Why this fork?

This fork of vitalets/google-translate-api contains several improvements with the primary change being it is written to support both the batch and single Google Translate endpoints, as well as any input request function. Additionally, there is optional language checking, and a list of supported languages that can best used. Many similar packages of the same endpoints either only use the single translate endpoint, which is quickly rate limited, or the batch translate endpoint which is sometimes innaccurate.

Install

npm install google-translate-api-x

Usage

From automatic language detection to English:

const translate = require('google-translate-api-x');
// Or of course
import translate from 'google-translate-api-x';
// Or deconstruct all the exposed variables as
import { translate, Translator, speak, singleTranslate, batchTranslate, languages, isSupported, getCode } from 'google-translate-api-x';
// or again
const { translate, Translator, speak, singleTranslate, batchTranslate, languages, isSupported, getCode } = require('google-translate-api-x');

const res = await translate('Ik spreek Engels', {to: 'en'});

console.log(res.text); //=> I speak English
console.log(res.from.language.iso);  //=> nl

If server returns Response code 403 (Forbidden) try set option client=gtx:

const res = await translate('Ik spreek Engels', { to: 'en', client: 'gtx' }).then(res => { ... });

Please note that maximum text length for translation call is 5000 characters. In case of longer text you should split it on chunks, see #20.

Autocorrect

From English to Dutch with a typo (autoCorrect):

const res = await translate('I spea Dutch!', { from: 'en', to: 'nl', autoCorrect: true });

console.log(res.from.text.didYouMean); // => false
console.log(res.from.text.autoCorrected); // => true
console.log(res.from.text.value); // => 'I [speak] Dutch!'

console.log(res.text); // => 'Ik spreek Nederlands!'

These reported values are often inaccurate and cannot be relied upon

Text-to-Speech (TTS)

A TTS request is made just like a translate request but using the speak method. The language spoken is the to language in options(or query params). The response is just a Base64 MP3 as a string.

import { speak } from 'google-translate-api-x';
import { writeFileSync } from 'fs';

const res = await speak('gata', {to: 'es'}); // => Base64 encoded mp3
writeFileSync('cat.mp3', res, {encoding:'base64'}); // Saves the mp3 to file

Single and Batch Endpoints

Single translate

The single translate endpoint is generally more accurate- and I suspect is a more advanced translation engine. Noticably it is more accurate(but still not completely accurate) when it comes to gendered words. Using the more accurate single translate endpoint requires setting forceBatch to false:

const res = await translate('cat', {to: 'es', forceBatch: false});

console.log(res.test); // => 'gato' instead of 'gata' which batch would return

And to guarantee it is used you can also pass fallbackBatch: false.

⚠️However, the single translate endpoint is much more likely to be ratelimited and have your request rejected than the batch translate endpoint!

Batch translate

Because of the risk of being ratelimited- the default endpoint and fallback endpoint for this package is the batch translate endpoint. It notably supports multiple queries in one request- so batch requests always go through the batch endpoint.

Persistent Options with Translator class

import { Translator } from 'google-translate-api-x';
const translator = new Translator({from: 'en', to: 'es', forceBatch: false, tld: 'es'});
const cat = await translator.translate('cat');
const dog = await translator.translate('dog');
const birds = await translator.translate(['owl', 'hawk']);

console.log(cat.text); // => 'gato'
console.log(dog.text); // => 'perro'
console.log([birds[0].text, birds[1].text]); // => '[ 'búho', 'halcón' ]'

Did you mean

Even with autocorrect disabled Google Translate will still attempt to correct errors, but will not use the correction for translation. However, it will update res.from.text.value with the corrected text:

const res = await translate('I spea Dutch!', { from: 'en', to: 'nl', autoCorrect: false });

console.log(res.from.text.didYouMean); // => true
console.log(res.from.text.autoCorrected); // => false
console.log(res.from.text.value); // => 'I [speak] Dutch!'

console.log(res.text); // => 'Ik speed Nederlands!'

Array and Object inputs (Batch Requests)

An array or object of inputs can be used to slightly lower the number of individual API calls:

const inputArray = [
  'I speak Dutch!',
  'Dutch is fun!',
  'And so is translating!'
];

const res = await translate(inputArray, { from: 'en', to: 'nl' });

console.log(res[0].text); // => 'Ik spreek Nederlands!'
console.log(res[1].text); // => 'Nederlands is leuk!'
console.log(res[2].text); // => 'En zo ook vertalen!'

and similarly with an object:

const inputObject = {
  name: 'Aidan Welch',
  fact: 'I\'m maintaining this project',
  birthMonth: 'February'
};

const res = await translate(inputObject, { from: 'en', to: 'ja' });

console.log(res.name.text); // => 'エイダンウェルチ'
console.log(res.fact.text); // => '私はこのプロジェクトを維持しています'
console.log(res.birthMonth.text); // => '2月'

If you use auto each input can even be in a different language!

Batch Request with Different Options (Option Query)

In Array and Object requests you can specify from, to, forceFrom, forceTo, and autoCorrect for each individual string to be translated by passing an object with the options set and the string as the text property. Like so:

const inputArray = [
  'I speak Dutch!',
  {text: 'I speak Japanese!', to: 'ja'},
  {text: '¡Hablo checo!', from: 'es', to: 'cs', forceTo: true, autoCorrect: true}
];

const res = await translate(inputArray, { from: 'en', to: 'nl' });

console.log(res[0].text); // => 'Ik spreek Nederlands!'
console.log(res[1].text); // => '私は日本語を話します!'
console.log(res[2].text); // => 'Mluvím česky!'

⚠️ You cannot pass a single translation by itself as an option query, it must either be passed as a string and use the options parameter, or passed wrapped by another array or object. This the translate function cannot differentiate between a batch query object and an option query object.

Using languages not supported in languages.js yet

If you know the ISO code used by Google Translate for a language and know it is supported but this API doesn't support it yet you can force it like so:

const res = await translate('Hello!', { from: 'en', to: 'as', forceTo: true });

console.log(res.text); // => 'নমস্কাৰ!'

forceFrom can be used in the same way.

You can also add languages in the code and use them in the translation:

translate = require('google-translate-api-x');
translate.languages['sr-Latn'] = 'Serbian Latin';

translate('translator', {to: 'sr-Latn'}).then(res => ...);

Proxy

Google Translate has request limits(supposedly for batch- and definitely for single translate). If too many requests are made, you can either end up with a 429 or a 503 error. You can use proxy to bypass them, the best way is to pass a proxy agent to fetch through requestOptions:

npm install https-proxy-agent
import HttpsProxyAgent from require('https-proxy-agent');
translate('Ik spreek Engels', {to: 'en', requestOptions: {
    agent: new HttpsProxyAgent('proxy-url-here');
  }
}).then(res => {
    // do something
});

Does it work from web page context?

It can, sort of. https://translate.google.com does not provide CORS http headers allowing access from other domains. However, this fork is written using Fetch and/or Axios, allowing contexts that don't request CORS access, such as a browser extension background script or React Native.

API

translate(input, options)

input

Type: string | string[] | {[key]: string} | {text: string, ...options}[] | { [key]: {text: string, ...options} }

The text to be translated, with optionally specific options for that text

Returns an object | object[] | {[key]: object}}:

Matches the structure of the input, so returns just the individual object if just a string is input, an array if an array is input, object with the same keys if an object is input. Regardless of that, each returned value will have this schema:

Note that res.from.text will only be returned if from.text.autoCorrected or from.text.didYouMean equals to true. In this case, it will have the corrections delimited with brackets ([ ]):

translate('I spea Dutch').then(res => {
    console.log(res.from.text.value);
    //=> I [speak] Dutch
}).catch(err => {
    console.error(err);
});

Otherwise, it will be an empty string ('').

speak(input, options)

input

Type: string | string[] | {[key]: string} | {text: string, ...options}[] | { [key]: {text: string, ...options} }

The text to be spoken, with optionally specific options for that text.

The to field of options is used for the language spoken in

Returns a string | string[] | {[key]: string}}:

The returned string is a Base64 encoded mp3.

Translator(options)

Returns a translator:

An object with these options as the default, that you can call the translate method on.

options

Type: object

from

Type: string Default: auto

The text language. Must be auto or one of the codes/names (not case sensitive) contained in languages.cjs.

to

Type: string Default: en

The language in which the text should be translated. Must be one of the codes/names (case sensitive!) contained in languages.cjs.

forceFrom

Type: boolean Default: false

Forces the translate function to use the from option as the iso code, without checking the languages list.

forceTo

Type: boolean Default: false

Forces the translate function to use the to option as the iso code, without checking the languages list.

forceBatch

Type: boolean Default: true

Forces the translate function to use the batch endpoint, which is less likely to be rate limited than the single endpoint.

fallbackBatch

type: boolean Default: true

Enables falling back to the batch endpoint if the single endpoint fails.

autoCorrect

Type: boolean Default: false

Autocorrects the inputs, and uses those corrections in the translation.

rejectOnPartialFail

Type: boolean Default: true

When true rejects whenever any translation in a batch fails- otherwise will just return null instead of that translation.

raw

Type: boolean Default: false

If true, the returned object will have a raw property with the raw response (string) from Google Translate.

requestFunction

Type: function Default: fetch

Function inputs should take (url, requestOptions) and mimick the response of the Fetch API with a res.text() and res.json() method.

client

Type: string Default: "t"

Query parameter client used in API calls. Can be t|gtx.

tld

Type: string Default: "com"

TLD for Google translate host to be used in API calls: https://translate.google.{tld}.

requestOptions

Type: object

The options used by the requestFunction. Must be in the style of fetchinit.

Related projects

License

MIT © Matheus Fernandes, forked and maintained by Aidan Welch.