15 languages out of the box
Russian, English, Spanish, German, French, Italian, Portuguese, Dutch, Polish, Czech, Turkish, Ukrainian, Japanese, Korean, Chinese — all at equal high quality.
🌐 Use case
Voice typing for bilinguals
If you work in two languages, switching layouts is slower than thinking itself. With voice, that problem disappears.
Sound familiar?
- Studying / working / communicating in two languages simultaneously
- Keyboard layout switching is slow and breaks your thinking
- Win+H behaves differently in RU and EN, with different quality
- Words from one language sneak into text in the other due to layout
What changes with AuroraWhisp
Switch in one click
Tray icon → pick a language — the app picks it up instantly. No restart. The hotkey stays the same; only the recognition model swaps.
Auto — the app figures out the language
Enable "Auto" mode in settings and the app detects the language from audio. Works well on phrases longer than 3 seconds. For short ones — pick the language explicitly.
No layout needed
With voice you do not think about QWERTY/Cyrillic. You said it — the app got it. No manual diacritics or special characters.
Real example: three languages daily
Hypothetical case: you work for a German company, your native language is Russian, English in international meetings. Three languages every day. Setup: Settings → Hotkeys: Ctrl+Space — RU (Slack with Russian colleagues), F9 — EN (meetings, Confluence), F10 — DE (internal mail, Jira). After two weeks, switching hotkeys becomes as automatic as Cmd+Tab between apps. No layouts to flip, no umlauts to remember.
Who actually needs this
Professions where bilingual work is the norm: translators (any pair), technical writers in international teams, immigrant freelancers (native + English / host country language), foreign-language teachers, content marketers across markets, IT engineers in distributed teams, diplomats and international-org staff, researchers in academic collaborations. If your work has three keyboard layouts and you switch them 50× a day — voice saves real hours per week.
Auto vs manual switching
Auto mode (app detects language from audio) works well when: phrases are longer than 3 seconds, and you stay in one language for that segment. Auto mode struggles when: short phrases (1-2 words), code-switching mid-sentence ("let's do a deploy via CI/CD"), similar-language mixing (Ukrainian / Russian, Spanish / Portuguese). Best approach: sustained bilingual work → manual switching by hotkey. Occasional sprinkles → leave Auto on, it copes most of the time.
What about accents?
Whisper models (Tiny / Base / Small / Medium / Large-v3) tolerate accents — they were trained on a huge variety of voices from YouTube and audiobooks. You can speak English with a strong Russian / German / Indian accent — Whisper Medium and Large handle it almost always. Sherpa Zipformer English is tuned for native English pronunciation and may slip on heavy accents. Advice for a bilingual with a noticeable accent in their second language: use Whisper Small / Medium instead of Sherpa models — they are less picky about pronunciation.
Reality check: code-switching
100% accuracy on mid-sentence code-switching — when you mix languages in one sentence ("let's open a pull request on the main branch and trigger CI") — nobody achieves yet. Not us, not Wispr Flow, not Google STT. Whisper Large-v3 is closest, but it still errs on terms. If that is critical for you — switch language explicitly before each foreign insertion, or edit after dictation. A fundamental current limitation of speech recognition technology.