No processing of your file contents takes place on any server. You need no data-processing agreement with gottrix to handle confidential or business documents.
Hosted in GermanyGlobal Content Delivery
Origin servers at Hetzner in Germany, delivered worldwide through a content delivery network. Your file contents are never transmitted in the process - they never leave your device.
An email to an English-speaking customer, a product description, a question for a landlord in Berlin: for short texts a solid draft translation that you polish yourself is often enough. This tool translates German to English and English to German with a neural translation model that runs entirely in your browser. The text is translated sentence by sentence, a progress bar shows how far it is, and paragraphs and line breaks are kept. You can copy the result or save it as a TXT file.
The models come from the OPUS-MT project of the University of Helsinki (Helsinki-NLP, Language Technology Research Group): opus-mt-de-en under the Apache 2.0 licence and opus-mt-en-de under the Creative Commons Attribution 4.0 licence (CC BY 4.0), trained on the freely available OPUS text collections. We use the quantised ONNX version by Xenova and serve it unchanged from this website. Each direction is about 112 MB, downloaded only when you start that direction for the first time; after that the model stays in your browser storage and the tool works offline too.
Honest about quality: this is machine translation with a compact model. Everyday sentences usually come out well; with technical language, idioms, wordplay, very long nested sentences or context that spans several sentences it gets noticeably weaker, and the model can get names, numbers or negations wrong. The quality is below that of large online translation services. So always check the translation before you send or publish it, especially for contracts, medical or legal texts.
Texts people want translated are often confidential: internal emails, job applications, draft contracts. Nothing is sent to a server here. Splitting into sentences, translation and output run via WebAssembly on your device, with no account and no storage. On a current computer a sentence takes roughly one second, longer on smartphones and older devices.
Specifications
Specifications
Input formats
Text input
Output format
TXT
Batch processing
No
Processing
Locally in your browser (WebAssembly)
File upload
None
In 3 steps
Paste your text into the field (up to 5000 characters).
Choose the direction and start; the first time, the model is downloaded once.
Check the translation, then copy it or save it as TXT.
Limitations: Only two directions: German to English and English to German; other languages are neither detected nor translated. Machine translation with a compact model: quality is below that of large online services, technical terms, idioms, names, numbers and negations can be wrong, and each sentence is translated on its own (no context across sentence boundaries). Check before any use. At most 5000 characters per run; very long sentences are split at commas or word boundaries. Formatting (bold, lists, tables) is lost; only paragraphs and line breaks are kept. Some special characters the model has no equivalent for (e.g. the vertical bar, emojis) are dropped. The first use of a direction downloads about 112 MB from this website; offline afterwards. Runs single-threaded in the browser, about one second per sentence on a current computer, much longer on phones.
FAQ
Is my text sent to a server?
No. Translation runs entirely in your browser. Only the model itself is downloaded from this website the first time; your text and the translation never leave your device.
Which languages are supported?
Exactly two directions: German to English and English to German. Each direction has its own model. No other language pairs are built in.
How good is the translation?
Usually usable to good for everyday sentences, but noticeably weaker than large online services, especially with technical language, idioms and long sentences. Use the result as a draft and check it before sending.
Why does the first run take so long?
The first time, the browser downloads the model for the chosen direction, about 112 MB. After that it stays in browser storage and starts quickly, even without internet. The other direction is only downloaded when you need it.
Whose model is this?
OPUS-MT by Helsinki-NLP (University of Helsinki): opus-mt-de-en under Apache 2.0, opus-mt-en-de under CC BY 4.0, in the ONNX version by Xenova. We have not changed the models, only made them available on this website.
Can I stop the translation?
Yes, with Cancel while it runs. Longer texts run sentence by sentence, and the progress bar shows how far it has got.