Skip to content

Home

Jump to bottom
tetherscript edited this page Feb 16, 2025 · 24 revisions

WinUITranslator

Translate your WinappSDK WinUI Apps using OpenAI, a local model or any other method you can dream up.


Important

  • Requires Visual Studio 2022 with .Net 8 and WinAppSDK 1.6
  • This Visual Studio 2022 solution contains a Translator app that translates your WinAppSDK WinUI project to the languages of your choice. It will scan your project, create the necessary translations, and update your Resource.resw files.
  • An OpenAI profile is included. You can translate an app for under $5 USD.
  • Local LMStudio profiles are included for deepseek-r1-distill-llama-8b, llama-3.2-1b-instruct, meta-llama-3.1-8b-instruct and qwen2.5-7b-instruct-1m models. Runs on a 8gb GPU, and for free! You can add new models by copying and modifiying an existing profile.
  • Try the scan and translated workflow with the included sample apps. It will just retrieve the already-translated cached translations and give you an idea of the workflow.
  • You can also code your own profile to call whatever translation service you want, including local and free models!
  • Run the included sample apps. They have already been translated from en-US to de-DE, fr-FR and ar-SA. You can modify these and re-translate them to see how it works.
  • Explore the included static class TLocalized and updated Resource.resw files.
  • You only need to include the TLocalized static class in your project. It allows dynamic text translations, like for a 'Loading {0}, please wait...' status message. It also allows the retrieval of non-translated resources, such as icon files, colors etc.
  • Your deployables do not require any third party packages or libraries. All the Translator-required code is contained in the included TLocalized.cs which is very simple, short and reviewable. In fact, if you don't use any dynamic translations, you don't even need to include that.
  • All translation settings and results are json/text files so you can include them in version control.
  • Trimming is supported.
  • Supports non-translateable resource items such as different icon and colors per culture.
  • You don't need to edit the .resw files. Ever.

Note

You use the Translator app to translate your target app.


Let's do a scan of the sample-packaged app. It will find anything that requires translation. scan1


Now we'll translate the scan results from en-US to ar-SA, de-DE, and fr-FR using OpenAI GPT4-o-mini. It has found that the en-US 'Connected' requires translation to ar-SA (Arabic), de-DE (Germani) and fr-FR (French). The translation result for de-DE is expanded to show the OpenAI API response, including reasoning. These translations are saved to a local cache file so that you don't need to re-translate these again.

translate1


Tip

Take a close look at the packaged and unpackaged sample apps. It helps clarify the obscure and somewhat undocumented process of setting the language and testing with RTL languages.

Clone this wiki locally