Auto-Translation
Dieser Inhalt ist noch nicht in deiner Sprache verfügbar.
In Crowdin uebersetzenAuto-translation helps speed up the translation process and ease the work of translators. It also allows you to save funds as part of translations can be automated. You can use and combine the following types of auto-translation:
- AI Auto-Translate - uses AI models to auto-translate the content.
- Machine Translation (MT) Auto-Translate – takes translations from supported machine translation engines (MTE) and applies them to the project content. You need to configure at least one of the translation engines to enable this feature.
- Translation Memory (TM) Auto-Translate – takes translations from the TM vault uploaded to the system and applies them to project content.
To apply auto-translation manually, follow these steps:
- Open your project and go to the Dashboard tab.
- Click Auto-Translation.

- In the auto-translate dialog, configure the following parameters:
- Auto-Translation via – select your preferred method from the drop-down list (AI, Machine Translation, or Translation Memory).
- AI Prompts (Specific to auto-translation via AI) – select the prompt that will guide the AI model’s behavior. Click the Manage prompts link to edit or add new prompts, or click to preview and edit the currently selected prompt.
- Custom instruction (Optional) (Specific to auto-translation via AI) – add any additional context or specific instructions to guide the AI (e.g., tone, domain). Your input will be included alongside the selected prompt.
- Translation engine (Specific to auto-translation via Machine Translation) – select the machine translation engine to use.
- Translation Memories (Specific to auto-translation via Translation Memory) – shows the translation memories from which the existing translations will be applied to your project content. Click Manage to configure the Translation Memories settings.
- Minimum match ratio (Specific to auto-translation via Translation Memory) – select either 100% or Perfect as the minimum similarity required for TM matches.
Read more about TM Match Types.
- Target languages – select the languages to auto-translate into. Use Search languages to quickly find the desired language.
- Expand the Advanced settings section to finetune the auto-translation process:
- Translate from – select the language to use as the source for auto-translation. By default, it’s preselected to the project’s source language, but you can choose any other project language. This provides greater flexibility and may improve translation quality by allowing you to use a more suitable existing translation (e.g., an adapted English version instead of the original Japanese) or translate directly between related languages (e.g., Spanish to Portuguese).
- Approve added translations (Specific to auto-translation via Translation Memory) – allows you to auto-approve translations added through auto-translation. Options include:
- With perfect match – approve only those with a 101% match.
- With perfect match (approved previously) – approve perfect matches only if they were already approved before.
- All (skip auto-substituted translations) – approve everything except translations improved by auto-substitution.
- All – approve all added translations.
- Filter by labels / Exclude labels – include or exclude specific strings based on their assigned labels.
Read more about Labels.
- In the Scope section, define which strings should be auto-translated:
- Strings to process – select Untranslated strings, Translated strings (Re-translation), or both.
- Re-translate only if last modified before – narrow down re-translation to strings whose existing translation was modified prior to a specific date (Anytime, 7 days ago, 30 days ago, 1 year ago, or Custom). Selecting Custom opens a date and time picker. This is especially useful for updating outdated translations after adding new contextual resources to your project, such as screenshots, style guides, or glossaries.
- When a translation already exists for this string – choose how to handle existing translations:
- Add translation alongside existing ones (default)
- Replace auto-generated translations (Keep human translations)
- Replace all existing translations
- Skip approved strings – prevents overwriting strings that already have approved translations.
- Allow duplicate translations – adds translations even if they duplicate existing ones. Useful when comparing AI prompts or MT engines in the Auto-Translation Accuracy report.
- Reset approval status so proofreaders can re-review – removes the approval status from any replaced translations, ensuring they can be reviewed and validated again.
- In the Files section (Specific to file-based projects) or Branches section (Specific to string-based projects), select the content you want to auto-translate. Use the search bar to quickly locate specific items.
- Click Auto-Translate.
Auto-translating Files and Folders via Context Menu
Section titled “Auto-translating Files and Folders via Context Menu”In addition to running the auto-translation from the project’s Dashboard, you can quickly auto-translate selected files or folders directly from the language page or Sources > Files using the context menu.
To auto-translate a single file or folder from the language page, follow these steps:
- Open your project and select the target language.
- Click on the needed file or folder and select Auto-Translate.
- In the auto-translate dialog, the target language and file/folder will already be selected. Configure the rest of the parameters as described in the Manual Auto-translation section.
- Click Auto-Translate.
To auto-translate one or multiple files or folders from the Sources > Files section, follow these steps:
- Open your project and go to Sources > Files.
- Click (or right-click) on the needed file or folder, or select multiple items using Ctrl , ⌘ , or Shift and then click (or right-click).
- Select Auto-Translate.
- In the auto-translate dialog, the files or folders you selected will already be checked. Configure the rest of the parameters as described in the Manual Auto-translation section.
- Click Auto-Translate.
The Auto-Translation queue allows you to trigger multiple manual auto-translations via TM, MT, or AI, each with different settings. After initiating an auto-translation, each instance enters a queue with a status of Pending.
Upon triggering a manual auto-translation, you will receive a notification indicating that the process has been added to the queue. To review all auto-translations currently in the queue, navigate to Tools > Auto-Translation queue in your project. Alternatively, click View Queue in the bottom-left corner of the auto-translate dialog to be redirected directly to this page.
Each entry in the Auto-Translation queue table includes the following details:
- Created at: Displays the date and time when the auto-translation was initiated.
- Type: Shows the method and specific engine or prompt used (e.g., AI, MT engine name).
- Priority: Indicates the execution priority of the task (low, normal, high). When the active auto-translation finishes, the system selects the pending job with the highest priority next. If multiple jobs share the same priority level, they are executed in the order they were created.
- Status: Pending, In progress, Completed, Failed, or Canceled. If a job fails, the system will automatically proceed to the next pending job in the queue without blocking it.
- Started at: The date and time the process began.
- Finished at: The date and time the process was completed.
You can use the Filters button to sort the queue by specific statuses, priorities, or types.
You can also manage pending jobs individually or in bulk. Select the checkboxes next to the relevant pending entries to change their priority or cancel them simultaneously using the Cancel Auto-Translation Jobs and priority buttons in the top toolbar.
For completed auto-translations, you can view a report detailing the results, including target languages, a list of processed files, the number of translations added, and the number of translations skipped with a breakdown by category (e.g., missing in TM, QA issue).
To repeat an auto-translation, click on the relevant entry and select Retry. This opens the auto-translation dialog with all the original settings pre-selected, allowing you to quickly launch the process again without manual reconfiguration. This is especially helpful if an auto-translation attempt fails or if you need to apply the same configuration again.
You can automate the Auto-Translation in the project Settings > Auto-Translate. In this case, each time you upload new content to your project, the system will automatically trigger the Auto-Translation and apply translations to untranslated content.
Read more about Auto-Translate New Content Automatically.
Handling QA Issues During Auto-translation via AI
Section titled “Handling QA Issues During Auto-translation via AI”If the AI provider returns translations with QA issues during auto-translation, Crowdin will collect the strings to which these faulty translations were applied and automatically reinitiate the auto-translation. The affected strings will be sent back to the AI provider along with details of the detected QA issues to ensure that the translations are corrected and meet quality standards.
Translation Memory Priority During Auto-translation via TM
Section titled “Translation Memory Priority During Auto-translation via TM”When using auto-translation via TM and the Share TMs option is enabled, Crowdin applies translations from translation memories based on a specific priority order. This priority ensures that translations come from the most relevant source, optimizing translation consistency and quality.
The order of priority is as follows:
- Default TM: Matches are first searched for in the project’s default TM.
- Assigned TMs: If no match is found in the default TM, the system continues searching through any TMs assigned specifically to the project.
- Shared TMs: If no match is found in the default or assigned TMs, translations are retrieved from shared TMs available to the project.
This sequence allows Crowdin to apply translations from the most relevant TM source first, ensuring that high-quality matches are prioritized when multiple TMs are available.