Automatic-Translations.md

June 7, 2026 ยท View on GitHub

Automated Translations

To speed up the localization process you can make use of automated translations.

Usage

Switch to the Translate tab to use this feature.

If the translation service needs credentials, enter them in the translation engine configuration. Follow the link to the translation service to learn about the service specifics. If you don't want to use a service, uncheck the check box next to the name.

After selecting source and target language, all entires that have no translation in the target language will be displayed and the translation runs in the background. When the translation is complete, the results can be reviewed or edited before accepting the result.

All results can be reviewed by opening the combo box in the target column. The results are ordered by the quality of the translation that the service reports. If you had already a translation of the same term in your project, the quality is the number of occurrences of this translation. Existing translations will have a higher quality rank than translations by a service.

Translations

Deepl

API Url

The API URL for Deepl is https://api.deepl.com/v2/translate, but for the free version it's https://api-free.deepl.com/v2/translate

Azure Open AI

Pros

  • Much more context aware
  • Higher quality translations, especially when fed with at least two different languages when translating to third/fourth/etc languages
  • Ability to use comments to guide the model
  • Better understanding of placeholders (such as {0}), seems to put them in better locations

Cons

  • Most models are slower, or much slower, than the other translation services
  • More expensive (depending on your token cost at Azure)

Configuration

  • Add a new Azure OpenAI resource using the portal or CLI.
  • Model availability may vary depending on e.g. region, check the model availability here
  • Deploy one of the supported models (see below).
  • Copy the API key, URL to the endpoint, deployment name, and model name into the settings of the translator.

Important: Understanding Model Name vs. Model Deployment Name

  • Model Name: This is the base model identifier used by OpenAI (e.g., "gpt-3.5-turbo", "gpt-4", "gpt-4o"). This is used internally for tokenization and determining which API endpoint to use.
  • Model Deployment Name: This is the custom name YOU chose when deploying the model in Azure (e.g., "my-gpt4-deployment"). This is what Azure uses to route your API requests.

A list of supported model names can be found here

Supported Models:

  • GPT-3.5 Models:

    • gpt-3.5-turbo-instruct (completion-based, legacy)
    • gpt-3.5-turbo (chat-based, recommended for GPT-3.5)
    • Note: You can also use gpt-35-turbo or gpt-35-turbo-instruct (Azure naming), which will be automatically mapped.

  • GPT-4 Models:

    • gpt-4 (chat-based)
    • gpt-4-turbo (chat-based, faster and more cost-effective)
    • gpt-4o (chat-based, latest multimodal model)
    • gpt-4o-mini (chat-based, smaller and faster variant)

Example Configuration:

  • If you deployed GPT-4 Turbo in Azure with deployment name "my-translation-model":
    • Model Name: gpt-4-turbo
    • Model Deployment Name: my-translation-model
    • Endpoint URL: https://your-resource-name.openai.azure.com

Additional Settings

  • You can add a custom prompt to your request to improve the translation quality or behavior, e.g. "preserve the html tags in the results"
  • You can include the comments in your resources in the prompt, to guide the model with additional hints about the context

Open AI

Pros

  • Same az Azure Open AI

Cons

  • Same as Azure Open AI but with a different pricing model
  • No batched translation support

Configuration

  • Go to here to sign up and get an API key from here
  • Copy the API key and a model of your choosing (e.g. "gpt-3.5-turbo"/"gpt-4-turbo") into the settings of the translator

Addtional Settings

  • You can add a custom prompt to your request to improve the translation quality or behavior, e.g. "preserve the html tags in the results"
  • You can include the comments in your resources in the prompt, to guide the model with additional hints about the context
  • If you need to use other LLM services compatible with the OpenAI API (such as DeepSeek), you can modify the URL to the corresponding service address (for example: https://api.deepseek.com). Since the internal token counting function supports a limited number of models, you may need to disable token counting (uncheck Max Tokens) to use other models properly.