Tilvar Beta

What does it do?

• Science-first translationKeeps terminology consistent, tries to preserve numbers, units, LaTeX and citations, and adds nothing to the text.
• Turkic languagesKazakh, Uzbek, Azerbaijani, Kyrgyz, Turkmen, Tatar, Crimean Tatar, Bashkir and more. Output is always written in the Turkey-Turkish alphabet.
• Text, documents, websitesPaste text, upload PDF / DOCX / TXT / HTML / SRT files or translate a website. Websites are only ever fetched through the Tor network.
• Editing workspaceSegment-by-segment editing, glossary, translation memory and TMX / XLIFF / CSV export.
• Team workSeveral people can work on the same translation: shared tasks and glossary, comments and live editing.
• On our own serverThe model is based on Google Gemma 4 and runs on our own server; your text is not sent to third-party translation services.
• APIUse it from browser extensions and your own tools through an OpenAI-compatible HTTP API (supporter tiers).

Languages

European languages, Chinese, Japanese, Korean, Vietnamese and the Turkic languages are supported. Mongolian, Tibetan, Manchu and Ancient Greek are experimental; almost any language is accepted as a source. The full list is inside the app.

Known limits

Translations are machine translations; always review important text. Quality varies more in low-resource languages.