Languages

Ditto has two language settings — one for what you see, one for what you say. Here's how they relate.

Two settings, one app

There are two places “language” appears in Ditto:

• App language (Settings → General → Interface language). Controls the text in menus, settings, the welcome window, the tray menu — everything Ditto draws on screen.
• Transcription language (Settings → Audio → Language). Tells Whisper which language to expect when you speak.

These are independent. You can have Ditto’s UI in English and transcribe in Spanish, or UI in Japanese and transcribe in French. Pick whatever combination matches how you actually work.

App language

Settings → General → Interface language offers nine options:

• Auto-detect — Ditto reads your Windows language and uses it if it’s one of the supported eight. Falls back to English if not.
• English
• Español
• Français
• Deutsch
• Italiano
• Português
• 日本語 (Japanese)
• 中文 (Chinese, Simplified)

Changes apply immediately, no restart needed. The pill, settings, tray menu, and welcome window all update on the spot.

Transcription language

Settings → Audio → Language controls what Whisper expects to hear. Same nine options as the UI: auto-detect or one of the eight languages.

For details on how this affects accuracy and latency, see the Audio input page. Quick summary:

• Auto-detect is convenient if you switch languages, but adds latency and can occasionally pick wrong.
• A fixed language is faster and more accurate when you only ever speak one.

Translate to English

There’s a third option that interacts with transcription language: Settings → Audio → Translate to English. When on, Whisper translates your speech into English regardless of which language you spoke.

This is independent of the UI language. You can have a Spanish UI and still get English text out of every transcription.

See the Audio input page for use cases and caveats.

Why two settings?

Practical reasons:

• People speak and read in different languages. A native Spanish speaker using Ditto in English at work might want UI in Spanish but transcribe in English.
• The model and the UI live in completely different parts of the app. Whisper.cpp doesn’t care about Ditto’s UI language; React doesn’t care about Whisper’s language hint. Keeping them separate avoids weird coupling.
• It mirrors what apps like macOS, Windows, and Office do: separate display language from input/dictation language.

Adding more languages

The eight current languages are the ones Whisper handles best at all model sizes and the ones with translated UI. Adding more requires:

• Translating ~150 strings in the UI for each new language.
• Confirming Whisper performs reasonably well at the model sizes Ditto ships.

If you’d like to contribute a translation, the strings live as JSON files in the repo at src/shared/locales/. Open an issue or PR on GitHub — every language file is about 200 keys.