DeepL API Too Expensive for Document Translation at Scale – Looking for Solutions

I have a question regarding the DeepL API billing model and how to handle the costs effectively.

I recently tested the DeepL API in a Power Automate flow. The flow itself works perfectly:

New file in SharePoint β†’ Get file content β†’ DeepL upload + translate β†’ Download result β†’ Save translated file back to SharePoint β†’ Send email with file link

Everything functions as expected.

The main problem: Unexpectedly high character usage

Even though the Word documents I tested contained only 150–200 visible characters, DeepL counts the internal XML structure of the .docx file. Because a Word file includes many hidden XML elements (styles, metadata, formatting, document structure, headers, etc.), the total character count quickly increases to 20,000–50,000 characters per file.

As a result:

My monthly limit of 350,000 characters was already consumed after only seven test translations.

This caused unexpected additional charges of around €12, even though the files were extremely small in terms of visible text.

My question for the community: How can this be solved?

I do not want to be in a situation where I pay for the €4.99 API plan but when someone in our company uploads a large document. for example, a 70-slide PPTX or a long Word file. we end up paying more than a normal DeepL Business subscription for individual users.

Is there any recommended solution for this such as :

  • A way to reduce character count before uploading?

  • A different pricing model for companies?

  • A bulk translation plan?

  • Any best practices to avoid translating unnecessary XML markup?

  • Or is the API simply not suitable for document translation at scale?

I would appreciate any tips, advice or alternative approaches ?

Regards,

Mohan

2 replies