The following sections define Translation Hub resources or concepts.


A portal is the user interface where authorized business users sign in and request translations. Portals are a Translation Hub resource and are created by a Google Cloud administrator in a Google Cloud project. Administrators add users to a portal to grant access. Portal users can't use or access the related Google Cloud project.

Translation Hub offers two portal tiers: basic or advanced. For more information, see Compare basic and advanced tiers.

Service account

A service account is used by an application, not a person, to authorize API calls to your project. Translation Hub uses a service account to make authorized calls to translate documents on behalf of portal users, who don't have Google Cloud credentials. Each portal has a unique service account, which administrators must add to their Google Cloud project. For more information about adding a portal service account, see Admin: Enable users to request translations.

Translation model

By default, Translation Hub uses the Google Neural Machine Translation (NMT) model to translate documents. Administrators can add custom translation models to portals (advanced tier only). Business users can then choose to use the NMT model or custom model to complete translations. A custom model, which is also known as an AutoML Translation model, is trained on your sentence pairs. Because the model is trained on your data, the model can produce translations that are tailored for your domain. To create and manage custom models, see the AutoML Translation documentation.


A glossary is a custom dictionary for consistently translating domain-specific terms and named entities. Each glossary entry is limited to 1024 UTF-8 bytes with a total file size limit of 10.4 million (10,485,760) UTF-8 bytes. Translation Hub automatically replaces matching terms with their associated target-language terms. For more information, see Admin: Enable users to request translations.

Translation memory

A translation memory (advanced tier only) stores segments as source-language and target-language segments. A translation memory is useful to set and maintain consistent translations at the sentence level. When you use a translation memory, Translation Hub automatically replaces matching source-language segments with their associated target-language segments.

To set up a translation memory, you can import your existing translations into Translation Hub or build your own by saving edits from post-editing. Translation Hub also lets you export your data as a TMX file. You might use this data, for example, to build custom models in AutoML Translation. For more information, see Translation memory.

Translation template

A translation template (basic tier only) is a predefined translation configuration that helps business users consistently translate documents.

In a template, you specify a source language, one or more target languages, and a glossary (optional) for each target language. When requesting a translation, business users can select a template instead of selecting the translation settings each time. Both administrators and business users can create translation templates. For more information, see Admin: Enable users to request translations.


Post-editing (advanced tier only) lets you bring in your own reviewers (translators) as part of the overall translation workflow. After business users complete a machine translation, they can send the results for post-editing. For more information, see Post-editing overview.

What's next