Every file goes through the same pipeline: parse the structure, apply your terminology, translate block by block with a frontier LLM, then reassemble the original — fonts, tables, figures and all.
The middle of the pipeline is shared across every format. Only the first and last stages — parsing and rebuilding — change depending on whether it's a PDF, a Word file, slides or a scan.
Detect the file type and extract every text block with its position and style into a structured map.
Match your scenario and glossary terms against each block before translation begins.
A frontier LLM translates each block in context, keeping inline formatting markers intact.
Translated text is placed back into the original layout, reassembling the file faithfully.
Compare source and translation side by side, then export the finished document.
Structure comes first. TransQ records where every block sits and how it's styled, swaps only the words, and rebuilds the original around them. The translation comes back looking like the document you started with.
Translation quality in a profession lives and dies on terms. Choose a scenario or upload your own glossary, and TransQ injects those term translations into the model, then verifies them after — so the same term is rendered the same way, end to end.
Every translation comes with a side-by-side preview: source on the left, translation on the right, in the document's real layout. Review it in full, and only pay when you're ready to download the finished file.
Most of what passes through TransQ is sensitive. The pipeline is built so files are encrypted, access is gated, and your content never trains a model.
The fastest way to judge fidelity is to try your own file. Preview is free — you only pay to download.