How to use
- Choose the source and target languages; the tool shows which model is needed and its licence.
- Press Download model once per language direction (cached afterwards).
- Paste text and press Translate; use Swap to translate the result back.
Worked example
“Where is the train station?” English → Spanish gives “¿Dónde está la estación de tren?”, and “The next train leaves in ten minutes.” gives “El siguiente tren sale en diez minutos.”
Supported formats and limits
| Input | Text up to 5,000 characters, TXT, Markdown, DOCX, SRT or VTT files |
|---|---|
| Output | Translated text (copy or download) |
| Limits | Each language direction is a separate model of about 100 to 130 MB, downloaded once from Hugging Face and cached. Pairs without a direct model are translated through English (two models). Your text never leaves this tab. |
| Engine | Opus-MT (Helsinki-NLP; Apache-2.0 or CC BY 4.0 per pair, ONNX by Xenova) via transformers.js in a Web Worker, sentence by sentence |
Limitations
- Machine translation can be wrong about idioms, names and tone, and each sentence is translated on its own; have important documents checked by a fluent speaker.
- Japanese, Korean, Polish, Turkish, Thai and Estonian can be translated into English only.
- Translating between two non-English languages goes through English and loses more meaning.
Questions
How good are the translations?
They are machine translations from small open models, made sentence by sentence. Idioms, names, tone and context across sentences can come out wrong. Have anything important checked by a fluent speaker.
What gets downloaded?
One Opus-MT model of about 100 to 130 MB per language direction from Hugging Face, cached afterwards. Licences are Apache-2.0 or CC BY 4.0 depending on the pair. Between two non-English languages, two models are used, through English, which loses more meaning.
Which files can I open?
TXT, Markdown, DOCX, SRT and VTT, up to 5,000 characters of text. The translation is copied or downloaded as a plain .txt file, not as a subtitle file, so check numbering and timing lines before reusing it.
Privacy
On-device model. Processing runs in this browser. The open-source model files are downloaded once from the model host (Hugging Face) and cached; your content is not uploaded.
- Hugging Face: Only after you press Download model: requests for the model files (one Opus-MT model of about 100 to 130 MB per language direction) from huggingface.co and its CDN, which see your IP address. Your text and files are never sent.
See the privacy policy for how toolsdocks handles data.