Skip to main content

Connect a Machine Translation (MT) Provider to wxrks

Connect and configure MT engines (Amazon, Google, Microsoft, DeepL, ODIN, TransPerfect, Intento, and more), control which one translates each language pair, and see how MT fits into the wxrks translation workflow.

Connect a Machine Translation (MT) Provider to wxrks

This guide covers Machine Translation (MT) in wxrks: which MT engines you can connect, how to configure your own credentials, and how MT fits into the rest of the translation workflow.

💡 Who is this for? This guide is for Account Admins who need to connect and manage Machine Translation provider credentials within the wxrks platform.


1. What Machine Translation does in wxrks

Machine Translation (MT) is the raw, segment-by-segment translation layer wxrks calls whenever it needs to fill in a target segment automatically — during a project's auto-translation, and as an in-Editor suggestion. It's a different system from the AI/LLM features described in Model Routing — how wxrks picks an AI model (Context-Sensitive translation, proofreading, Sous Chef, and the rest): MT engines translate one segment at a time, while the AI providers configured in Model Management use the wider context of the file. Many projects use both together — see How MT fits into the translation workflow below.

Every wxrks account already has an MT engine active by default, labeled in the product as wxrks (Powered by Microsoft) and marked Included in subscription — enabled automatically, with no setup or credentials required. You don't need to configure anything to start using MT — the steps below are for connecting your own credentials from an external provider, either to replace the default or to use for specific language pairs.


2. Supported MT providers

You can link credentials from any of the following providers to your wxrks account:

Provider

Best for

What you need

Amazon Translate

Teams already using AWS

AWS Access Key ID + Secret Access Key

Google Cloud Translation API

Broad language coverage

A Google API key, or a Service Account JSON key

Microsoft Azure Translator

Enterprise Azure deployments

Subscription key, endpoint, and region

DeepL API

High-quality European-language pairs, with tone control

A DeepL API key (Pro or Free)

Google AutoML API

A custom-trained Google translation model

Service Account JSON + your AutoML model ID

ODIN

Teams with an existing ODIN account

Project ID, API key, and API secret

TransPerfect

TransPerfect GlobalLink customers

A TransPerfect subscription key

Intento

Routing to multiple MT engines through one account

An Intento API key

These credentials are obtained directly from each provider — wxrks does not issue or resell them. Consult the provider's own documentation or support for instructions on generating the correct credentials.


3. Where to configure MT

Go to Settings → Translation Settings → MT Settings.

MT Settings page with the built-in wxrks (Powered by Microsoft) engine and connected providers listed

4. Adding an MT configuration — step by step

On the MT Settings page, fill in a new configuration:

Add MT configuration form with Config name, Provider, Set as default, Set a language pair, and Credentials fields highlighted
  1. Config name — a label so you recognize this configuration later.

  2. Provider — pick one of the providers above from the dropdown. The dropdown lists them by their short id rather than a full name — aws, google, microsoft, deepl, automl, odin, transperfect, and intento — matching the table in section 2 in that order.

Provider dropdown open, listing aws, google, microsoft, deepl, automl, odin, transperfect, and intento
  1. Credentials — paste the provider's credentials as JSON, or use Upload Json to load a credentials file (this field doesn't apply to the built-in wxrks engine). If you pick deepl, note that wxrks currently supports only the DeepL API (Pro or Free) — the standalone DeepL Pro plan without API access is not supported.

  2. Set as default — makes this configuration the one used whenever no more specific language pair or level takes priority (see section 7).

  3. Set a language pair — optionally restrict this configuration to one specific source → target pair instead of (or in addition to) using it as a default.

Click Save. wxrks validates the credentials immediately — a warning toast means the credentials couldn't be verified, but the configuration is still saved so you can edit and retry.


5. Testing, viewing, editing, and removing a configuration

Every configuration in the list (shown in the screenshot in section 3) carries its own row of controls: a radio button to make it the account's active default, a refresh icon that opens Test credentials — pick a source and target language, type a short sample text, and run it to confirm the credentials actually translate before you rely on them in a real project — an eye icon that shows the configuration's raw credentials JSON with a copy button, a pencil icon to reopen it for editing, and a trash icon to delete it after a confirmation prompt. Credential fields are left blank when editing — leave them blank to keep the existing values, or fill them in to replace them. The built-in wxrks (Powered by Microsoft) row only shows an info icon instead, since it's fixed and can't be modified or removed.


6. Credential details by provider

The exact fields wxrks expects for each provider:

  • Amazon Translate — your AWS Access Key ID and Secret Access Key:

    {
    "access_key_id": "YOUR_ACCESS_KEY",
    "secret_access_key": "YOUR_SECRET_ACCESS_KEY"
    }
  • Google Cloud Translation API — either a simple API key:

    {
    "api_key": "YOUR_API_KEY"
    }

    or a Service Account key generated in the Google Cloud Console (the method Google itself recommends, since it doesn't rely on a single long-lived key):

    {
    "type": "service_account",
    "project_id": "xxxxxx",
    "private_key_id": "xxxxxx",
    "private_key": "xxxxxx",
    "client_email": "xxxxxx",
    "client_id": "xxxxxx",
    "auth_uri": "xxxxxx",
    "token_uri": "xxxxxx",
    "auth_provider_x509_cert_url": "xxxxxx",
    "client_x509_cert_url": "xxxxxx"
    }
  • Microsoft Azure Translator — the subscription details from your Azure resource:

    {
    "endpoint": "YOUR_ENDPOINT",
    "subscription_key": "YOUR_SUBSCRIPTION_KEY",
    "location": "YOUR_REGION"
    }
  • DeepL API — your DeepL key, with an optional formality setting for the translation's tone:

    {
    "key": "YOUR_KEY_HERE",
    "formality": "more"
    }

    Formality options are default, more (formal), less (informal), prefer_more, and prefer_less (the last two fall back to default formality for a target language that doesn't support it). DeepL only honors formality for a subset of target languages — check DeepL's own API documentation for the current list, since DeepL adds language support independently of wxrks.

  • Google AutoML API — the same Service Account JSON as above, plus a model field pointing at your trained model:

    {
    ...service account fields...,
    "model": "projects/PROJECT_ID/locations/LOCATION/models/MODEL_ID"
    }
  • ODIN — your ODIN project and API credentials:

    {
    "project-id": "YOUR_PROJECT_ID",
    "X-API-KEY": "YOUR_API_KEY",
    "X-API-SECRET": "YOUR_API_SECRET"
    }
  • TransPerfect — your subscription key, plus optional routing details if your TransPerfect account uses them:

    {
    "subscription_key": "YOUR_SUBSCRIPTION_KEY",
    "domain": "YOUR_DOMAIN",
    "tmID": "YOUR_TM_ID",
    "engineId": "YOUR_ENGINE_ID"
    }
  • Intento — your Intento API key. Intento itself routes to one or more underlying MT engines, so you can optionally tell it which provider and routing strategy to use — check your Intento account for the exact values it expects:

    {
    "intentoApiKey": "YOUR_API_KEY",
    "intentoProvider": "YOUR_INTENTO_PROVIDER",
    "intentoRouting": "YOUR_INTENTO_ROUTING"
    }

Regardless of the provider, credentials are stored encrypted and are not read directly by wxrks staff — the system references them through a config ID generated when you save the credentials. You can use that same configuration to remove the credentials at any time.


7. Applying a configuration at the Account, Organization, or Org Unit level

A saved MT configuration can be applied at three levels, from broadest to most specific:

  • Account — applies to every project in your account, unless overridden below.

  • Organization — applies to projects under one Organization. Open the Organization and go to its MT tab.

  • Org Unit — applies to projects under one Org Unit. Open the Org Unit and go to its MT tab.

Org Unit page MT tab, showing its own default provider dropdown and language-pair list, falling back to the Account default when none is set

Each level works the same way: set a default configuration for that level, and optionally add language pairs that point to a different configuration than the default. For example, you could use Google for en-US → pt-BR and Amazon Translate for en-US → es, as long as both providers are already configured on your account.

When wxrks looks up which MT configuration to use for a segment, it checks from most to least specific:

  1. A language-pair override at the Org Unit level.

  2. The Org Unit's default configuration.

  3. A language-pair override at the Organization level.

  4. The Organization's default configuration.

  5. A language-pair override at the Account level.

  6. The Account's default configuration.

  7. If none of the above are set, the built-in wxrks (Powered by Microsoft) engine.


8. How MT fits into the translation workflow

Connecting a provider here doesn't translate anything by itself — a project has to actually call on MT. That's controlled per-project in Project's Translation Settings, under Auto Translate From: a project can auto-translate using Context-Sensitive AI translation, Translation Memory, Machine Translation, or a combination. When more than one is enabled, wxrks prioritizes Context-Sensitive first, then Translation Memory, then Machine Translation.

Separately, a project setting called Show Machine Translation suggestion in Editor surfaces an MT suggestion in the Editor's Suggestions tab even for segments that weren't auto-translated by MT — useful when you want MT available as a reference without letting it fill segments automatically.

MT and the AI/LLM providers in Model Routing — how wxrks picks an AI model are configured independently of each other — connecting an MT provider here has no effect on which AI model powers Context-Sensitive translation, proofreading, or Sous Chef, and vice versa.


9. Troubleshooting

"I saved a provider but got a credential warning." wxrks tried to validate the credentials against the provider and couldn't. Double-check the field names and values match the formats in section 6 exactly, then use the Test credentials action (section 5) to confirm — the configuration is still saved, so you can edit it and re-enter the credentials without starting over.

"Can I delete or recreate the wxrks (Powered by Microsoft) row?" No — it's provisioned automatically for every account and can't be manually created, modified, or removed. You can still add other providers alongside it and set one of them as the new default.

For anything else, search the wxrks Community or contact [email protected].


Did this answer your question?