Skip to content

Your first translation

A quick end-to-end run — under 5 minutes once setup is done.

Before you start

You need installation finished and an LLM API key configured. The free Google Gemini tier is enough for a first try.

Translate a Word document

  1. Launch the desktop app:

    uv run python -m src.main

  2. Click Translate Document in the left sidebar.

  3. Drag any .docx file into the drop zone — or click Browse to pick one.

  4. The file appears in the queue. Pick a target language at the top:

    • Source: Auto-detect (default — usually correct)
    • Target: e.g. French, Vietnamese, Japanese, Chinese (Simplified)

  5. Click Translate (or press Ctrl+Enter).

  6. Watch the progress bar in the history table at the bottom of the page. When it reaches 100%, click Open on the row to open the translated file — saved next to the original with a _translated_<src>_<tgt> suffix.

What just happened

  • Your .docx got cloned into a per-task storage folder so the original is untouched.
  • The text was extracted, batched into LLM-friendly chunks, translated, then re-injected into the document with all formatting preserved (bold, italic, fonts, colours, headers, footnotes, hyperlinks…).
  • A history entry was written to a SQLite database so you can re-open, re-run, or re-translate the file later.

Try the quick wins next

Pop into Translate Text in the sidebar. Paste anything, pick a target, hit Enter. Streaming output, language-swap (Ctrl+L), edit mode, TTS playback.

Generate Subtitle — drop an .mp4. You'll get an .srt back in the source language. (To translate and dub the video, use the Dubbing page instead.)

Live Translation — pick microphone or system audio, pick a target, Start. A floating overlay window shows captions in real time.

Where to next

  • See the feature index for what each page does.
  • Wire up more providers (custom endpoints, ElevenLabs, Soniox, Google Cloud).
  • Try the CLI for batch / scripted runs.