Edit online

AI Translator Add-on

The Oxygen AI Translator add-on is a powerful Oxygen XML extension that provides AI-powered translation capabilities for XLIFF (XML Localization Interchange File Format) files. This add-on streamlines the translation workflow by leveraging advanced AI models to produce high-quality translations while maintaining the structural integrity of your XML content.

What is a Translation Task?

When working with the Oxygen AI Translator add-on, a translation task represents a complete translation workflow that leverages the AI connectors and models configured in the Oxygen AI Positron to translate XLIFF translation units. Each task operates on specific XLIFF files and applies the selected AI translation configuration to process the content systematically.

Key Features

  • AI-Powered Translation - Translation tasks utilize the AI connectors and models defined in the Oxygen AI Positron, ensuring consistent configuration across Oxygen tools. This integration provides access to various AI engines with the flexibility to choose the most appropriate model for your translation requirements.
  • Quality Assurance through Introspection - Each translation task includes a built-in introspection mechanism that verifies the accuracy and quality of translations. This quality control system analyzes translated content to ensure it maintains the integrity of the original formatting and structure.
  • Retry Translation System - When introspection identifies potential issues with a translation, the system can automatically initiate additional translation attempts using different models. This progressive approach allows the use of more powerful models to ensure optimal translation quality.
  • Parallel Task Execution - The add-on supports running multiple translation tasks simultaneously, enabling improved productivity and faster processing of large translation projects.
  • Configurable Translation Settings - Translation configurations are fully editable through the preferences interface, allowing you to customize AI models, retry strategies, and introspection parameters.
Edit online

Installation

Prerequisites

The Oxygen AI Positron add-on must be installed for the Oxygen AI Translator to function properly. The AI connectors defined in Oxygen AI Positron can also be used for AI translations, providing a unified configuration experience across Oxygen tools.

Quick Installation

You can drag the following Install button and drop it into the main editor in Oxygen to quickly initiate the installation process:

Install

Manual Installation

To manually install the add-on, follow these instructions:

  1. Go to Help > Install new add-ons to open an add-on selection dialog box.
  2. Enter or paste http://www.oxygenxml.com/InstData/Addons/default/updateSite.xml in the Show add-ons from field.
  3. Select the Oxygen AI Translator add-on and click Next.
  4. Read the end-user license agreement. Then select the I accept all terms of the end-user license agreement option and click Install.
  5. Restart the application.

Result: The AI Translator view is now available and can be selected from the Window > Show View menu.

Edit online

AI Translator View

The AI Translator view is the main interface for managing translation tasks in the Oxygen XML application. This view provides a comprehensive task management system that allows you to create, monitor, and control translation jobs with real-time status updates and detailed progress tracking.

Figure 1. AI Translator View

Main Toolbar

New task button
Opens the New Translation Task dialog box where you can create a new task.
Settings drop-down
Provides access to the following options:
Preferences
Opens the Preferences dialog box where you can configure various options for the AI Translator.
Clear cache
Clears the translation cache.

Task List Panel

The main panel displays all translation tasks and includes the following information for each task:
  • Task Name - Descriptive name of the translation job.
  • Progress Bar - Visual indicator that shows the completion percentage and current file being processed.
  • Status Badge - Color-coded status indicator (Queued, Running, Stopped, Completed).
  • Task Details - Creation date, progress information, and duration.

The following actions are possible in the main panel, depending on the status:

Resume Task
Runs the task execution.
Stop Task
Runs the task execution.
Delete Task
Removes the task from the list.
Save Translated XLIFF (available for Completed tasks)
Saves the current translation work as an XLIFF file on your computer.

Context Menu Actions

The following contextual menu actions are available when right-clicking a task box:



Resume Task
Runs the task execution.
Stop Task
Runs the task execution.
Delete Task
Removes the task from the list.
Duplicate Task
Creates a new task with the same information as the selected task (user instructions, translation memory, XLIFF files, and the AI model configuration used).
Retry Fails Translations (available for Completed tasks)
Allows you to create a new task with different settings and instructions to retry translating the failed parts.
Save Log File
Saves a log file with detailed information about the task execution, current status, and any problems encountered.
Save Translated XLIFF (available for Completed tasks)
Saves the current translation work as an XLIFF file on your computer.
Save Intermediate Translated XLIFF (available for Stopped or Failed tasks)
Saves the current translation work as an XLIFF file on your computer after some parts have been translated.

Task Status States

Translation tasks progress through various states during their lifecycle:

  • Queued - Task is created and waiting to be processed.
    • Shows the creation timestamp.
    • Task is in the processing queue.
  • Running - Task is actively being processed.
    • Displays real-time progress (e.g. "15/400 translation units").
    • Shows the current file being processed.
    • Provides the estimated time remaining.
    • Updates the elapsed processing time.
  • Stopped- Task has been paused or interrupted.
    • Shows the last update timestamp.
    • Can be resumed using the play button.
    • Preserves the current progress.
  • Completed - Task has finished successfully.
    • Shows the completion timestamp and total duration.
    • Displays the final statistics (e.g. "400/400 translation units").
    • Task remains in the history for reference purposes.

Task Management Features

The comprehensive task management system ensures efficient handling of translation workflows while providing full visibility into the translation process.

Progress Tracking
  • Real-time progress bars that show the completion percentage.
  • The current file being processed.
  • The translation units completed vs. total.
  • The estimated time remaining for running tasks.
Task Control
  • Start, pause, and resume individual tasks.
  • Delete completed or unwanted tasks.
  • Queue multiple tasks for sequential processing.
Status Monitoring
  • Visual status indicators for quick task state identification.
  • Detailed progress information for each task.
  • Historical record of completed tasks with performance metrics.

Caching Mechanism

The AI Translator add-on includes a caching system that stores AI translation responses locally to optimize performance, reduce costs, and improve the overall translation experience.

Why Caching is Beneficial

  • Cost Reduction - By caching AI engine responses, the system significantly reduces API calls to external services such as OpenAI, resulting in lower usage costs for repeated translation requests.
  • Performance Improvement - Cached translations are retrieved instantly from local storage, eliminating network latency and API response times for previously translated content.
  • Consistency - The caching mechanism ensures that identical translation requests always return the same results, providing consistent and reproducible translations across multiple runs.

Enabling and Configuring Caching

The caching configuration is managed through the AI Translator Preferences dialog box. You can access preferences using the following methods:
  • From the Oxygen AI Translator view - Select Preferences from the Settings drop-down.
  • From the Oxygen menu - Go to Options > Preferences > Plugins > AI Translator.

Clearing the Cache

You can clear the cache from the Oxygen AI Translator view by using the Clear cache button from the Settings drop-down.

Clearing cache might be useful after significant changes to AI model configurations or when experiencing inconsistent translation results.

Edit online

Creating a New Translation Task

Figure 2. New Translation Task Dialog Box


To create a new translation task in the AI Translator:

  1. Open the Oxygen AI Translator view (go to Window > Show View > Oxygen AI Translator if it is not visible).
  2. Click the New task button in the AI Translator toolbar to open the New Translation Task dialog box.
  3. Configure the task settings:
    • Task name - Enter a descriptive name for your translation task (e.g. My First Task).
    • Configuration - Select the desired AI translation configuration from the drop-down menu (e.g. "Translate with GPT 4.1").
    • Custom prompt instructions file (Optional) - Specify a text/markdown file with custom instructions to refine AI translations. Use this to define scope, audience, style, or guidelines for the translation.
  4. Specify files:
    • XLIFF file - Browse and select the XLIFF file you want to translate. Currently only XLIFF version 1.2 files are supported and must be valid to create the translation task.
    • Memory file – (Optional) Specify a translation memory file in TMX format to improve translation consistency and leverage previously translated segments. The TMX file must be valid according to the TMX 1.4b Specification. Using a valid TMX file allows the AI Translator to reuse existing translations for matching segments, ensuring greater accuracy, consistency, and efficiency in your translation workflow.
  5. Click OK to start the translation process, or Cancel to abort.

Once created, your translation task will appear in the Oxygen AI Translator view with status indicators that show the progress (e.g. Running, Stopped, or Completed) and processing time information.

Edit online

Configuration and Settings

The AI Translator add-on provides comprehensive configuration options in the Preferences dialog box, allowing you to customize translation behavior, AI model settings, caching, and retry mechanisms.

Figure 3. Preferences Dialog Box


You can access preferences using the following methods:
  • From the AI Translator view - Select Preferences from the Settings drop-down.
  • From the Oxygen menu - Go to Options > Preferences > Plugins > Oxygen AI Translator.

The Oxygen AI Translator preferences page contains the following options:

Maximum Parallel Translation Tasks
Sets the maximum number of translation tasks that can run at the same time. Increasing this value can speed up the translation process by allowing more tasks to be processed in parallel, depending on your system resources and AI service limits.
Recommendation:
Adjust this value based on your computer’s performance and the capabilities of your AI provider. A higher value may improve throughput, but setting it too high could lead to resource contention or rate limiting.
Enable Caching

Enables or disables the local caching of AI engine responses. When enabled, messages exchanged with AI engines are cached locally for reuse and faster translations.

Size on Disk (MB)
Specifies the cache size limit (default is 100 MB).
Time-to-Live (Days)
The cache entry expiration time (default is 7 days). The drop-down offers the following options: 3, 5, 7, 15, 31, 90, 180, or 365 days. A cache entry expires after the configured time-to-live period.
Cache Configuration Options
  • Checkbox - Toggle to enable/disable the caching system.
  • Default - Enabled (recommended for optimal performance).
Size on Disk
  • Setting - Configure maximum cache size in MB.
  • Default - 100 MB.
  • Range - Adjustable based on available disk space and translation volume.
  • Recommendation - Set to 100-500 MB for typical usage, higher for large-scale translation projects.
Time-to-Live/Expiration
  • Setting - Configure cache entry expiration time.
  • Default - 7 days.
  • Options - 3, 5, 7, 15, 31, 90, 180, 365 days.
  • Recommendation - 7-15 days for active projects, longer periods for reference translations.
AI Configurations
Each AI configuration defines how translations are performed using the Oxygen AI Positron connectors and selected AI models. You can configure the primary translation model, fallback (retry) models, their order, and enable quality checks for chosen translation units.
The configuration table contains the following columns:
  • Name - Descriptive name for the configuration (e.g., "Translate with GPT 4.1").
  • AI model - The AI model used for translation (e.g., "gpt-4.1", "gpt-5", "claude-sonnet-4-5", "gemini-2.5-flash").
Default AI configurations:
  • Translate with GPT 4.1 (gpt-4.1)
  • Translate with GPT 5 (gpt-5)
  • Translate with Claude 4.5 Sonnet (claude-sonnet-4-5)
  • Translate with Gemini 2.5 Flash (gemini-2.5-flash)
The following buttons are available under the table:
  • Add - Create a new translation configuration.
  • Edit - Modify an existing configuration.
  • Delete - Remove a configuration.
Edit online

Translation Configuration Dialog Box

Figure 4. Translation Configuration Dialog Box


The Add/Edit Configuration dialog box defines how translations are performed using an Oxygen AI Positron connector and AI models. You can configure retry models, quality checks, and related behavior.

General Settings

  • Configuration name - Enter a custom name for this configuration (e.g., "Fast Config").
  • Translation model - Select the primary AI model to use for translation (e.g., "GPT-4.1 Nano").

Quality Checks

  • Enable quality checks & retries - When checked, the system will run additional checks to verify translation quality. If a check fails, the system can retry the translation using alternative models. This may incur additional costs due to increased token usage.
  • Apply to - Choose which translation units are subject to quality checks and retries:
    • All translation units - Quality checks and retries are applied to every translation unit.
    • Only for translation units with inline markups - Quality checks and retries are performed only for translation units that contain inline tags or formatting (such as bold, italics, or links).
  • Check quality model - Select the AI model used for quality checks (e.g., "GPT-4.1 Mini").

Retry Models

If translation or quality checks fail, the system will retry using the next model in the list. You can select up to three fallback models, in order:

  • 1st retry - First fallback model to use (e.g., "GPT-4.1 Mini").
  • 2nd retry - Second fallback model to use (e.g., "GPT-4.1").
  • 3rd retry - Third fallback model to use (e.g., "GPT-5").

Click OK to save your configuration or Cancel to discard changes.

Edit online

Best Practices

Caching Configuration

  • Enable caching for faster repeat translations and reduced API costs.
  • Set a size limit based on available disk space and translation volume.
  • Balance the expiration time between freshness and reuse (7-15 days recommended).

Model Selection

  • Choose the Primary model based on translation quality requirements and cost considerations.
  • Configure a mix of Retry models for fallback scenarios.
  • Use Introspection for complex documents that require a high level of accuracy.

Retry Strategy

  • A number of 3 max retries is typically sufficient for most scenarios.
  • For the model sequence, start with cost-effective models, then escalate to premium models for retries.
  • Enable complex units only for documents with complex inline formatting.

This configuration system provides flexibility to optimize translation quality, performance, and cost based on your specific requirements and available AI models.