If you work in multiple languages, standard dictation apps make you manually switch languages every time. Whisperer handles multilingual workflows natively — with per-app language profiles, AI-powered translation, and over 100 supported languages.
Setting Your Dictation Language#
The most important step for accuracy: explicitly select your dictation language rather than relying on auto-detection.
Auto-detection works but can misidentify languages, especially for:
- Short utterances (fewer words = less data for the model)
- Languages with similar phonetics (e.g., Spanish and Portuguese)
- Code-switching (mixing languages in one utterance)
Tip: Explicit language selection gives 10-15% better accuracy than auto-detection for most languages.
Per-App Language Profiles#
With Whisperer's Pro Pack, you can assign a different dictation language to each application:
- VS Code / Cursor: English (for code and comments)
- Pages / Word: French (for documents)
- Mail: German (for correspondence)
- Slack: English (for team chat)
When you switch between apps, Whisperer automatically switches the dictation language. No manual toggling needed.
AI Translation Mode#
Whisperer's AI post-processing includes a Translate mode that converts your dictation to a different language entirely offline:
- Enable AI Post-Processing in Settings
- Select "Translate" mode
- Set your target language
- Dictate in your native language — the output appears in the target language
This is particularly useful for:
- Writing emails in a language you can speak but type slowly in
- Quick translations without opening a separate app
- Maintaining consistent terminology with your personal dictionary
Language Support by Backend#
Not all backends support the same languages:
| Backend | Languages | Best For |
|---|---|---|
| Whisper | 99+ | Maximum language coverage |
| Parakeet v3 | 25 | Fast inference for top languages |
| Parakeet v2 | English only | Maximum English speed |
| Apple Speech | System languages | Native macOS integration |
If you need a less common language, Whisper is the only backend with full coverage. For top-25 languages, Parakeet offers the fastest inference.
Tips for Non-English Accuracy#
- Use the right model size. Larger Whisper models (Medium, Large V3) perform significantly better on non-English languages than smaller models
- Add prompt words for proper nouns, place names, and technical terms in your language
- Set up dictionary corrections for words the model consistently gets wrong
- Use explicit language selection rather than auto-detection
Getting Started#
- Download Whisperer
- Go to Settings and select your language
- Optionally set up per-app profiles for multilingual workflows
- Explore the full multilingual features