Speak in your native language, get accurate text, and translate it anywhere. Voice dictation makes multilingual work effortless.
Working across languages is a daily reality for millions of professionals. Whether you are a translator handling client documents, a remote worker communicating with international teams, a student studying foreign languages, or a business owner serving customers in multiple markets, the ability to quickly move between languages is essential. The traditional workflow — typing in one language, copying to a translator, reviewing the output — is slow and error-prone, especially when the source language is not your keyboard's primary layout.
Voice dictation transforms this workflow. Instead of hunting for accented characters or switching keyboard layouts, you simply speak in whatever language feels natural. Steno transcribes your speech accurately using Groq Whisper, which supports over 50 languages. The transcribed text can then be translated using your preferred service. This voice-to-text-to-translation pipeline is faster, more accurate, and far less frustrating than typing in a non-native keyboard layout.
The key insight is that translation is a two-step process: first you need accurate source text, then you need accurate translation. Most translation errors come from poor source text — typos, missing words, garbled sentences. Voice dictation produces cleaner source text because speaking is more natural than typing, especially in languages where the keyboard layout does not match the language's character set.
This workflow pairs Steno's accurate transcription with dedicated translation tools like DeepL, Google Translate, or professional CAT (Computer-Assisted Translation) tools. By separating transcription from translation, you get better results in both steps.
1 Select your transcription language in Steno
Open Steno's settings from the menu bar icon and select the language you will be speaking in. Steno uses Groq Whisper, which supports over 50 languages including English, Spanish, French, German, Italian, Portuguese, Japanese, Chinese (Mandarin), Korean, Hindi, Arabic, Russian, and many more. Selecting the correct language ensures optimal transcription accuracy for your speech patterns and vocabulary.
2 Open your working document and translation tool
Set up your workspace with two windows side by side. On one side, have the document or text field where you will dictate. On the other, have your translation service open — DeepL, Google Translate, or whatever tool you prefer. This split-screen setup minimizes the time spent switching between apps. On macOS, you can use Split View or a window manager to arrange this efficiently.
3 Dictate your source text
Place your cursor in your working document. Hold the Steno hotkey and speak clearly in your chosen language. Speak at a natural pace. There is no need to slow down or over-pronounce words. Whisper handles natural speech patterns, contractions, and colloquialisms well. For best results, speak in complete sentences rather than isolated words or fragments, as the context helps the model produce more accurate transcription.
4 Review the transcription
After releasing the hotkey, quickly review the transcribed text. Look for any words that the model might have misheard, especially proper nouns, brand names, or domain-specific terminology. Fix any errors in the transcription before moving to translation, because errors in the source text propagate and amplify in the translation output.
5 Translate the text
Select the transcribed text, copy it, and paste it into your translation tool. Alternatively, if you are working directly in a tool that has built-in translation (like some versions of Google Docs), you can access the translation feature directly. Review the translation for accuracy, paying particular attention to idiomatic expressions, cultural references, and technical terms that may not translate directly.
6 Refine for context
Machine translation has improved dramatically but still struggles with context-dependent meaning. Read through the translated text and adjust any phrases that sound unnatural or miss the intended meaning. For professional translation work, this human review step is essential and is where your bilingual expertise adds the most value.
"Estimado equipo, quería informarles que la reunión del viernes ha sido reprogramada para el lunes a las diez de la mañana. Por favor confirmen su asistencia antes del jueves."
Transcription output: Estimado equipo, quería informarles que la reunión del viernes ha sido reprogramada para el lunes a las diez de la mañana. Por favor confirmen su asistencia antes del jueves.
After translation: Dear team, I wanted to inform you that Friday's meeting has been rescheduled for Monday at ten in the morning. Please confirm your attendance before Thursday.
"L'API accepte les requêtes POST avec un corps JSON contenant l'identifiant de l'utilisateur et le jeton d'authentification. La réponse inclut le statut de la transaction et un code de confirmation."
Transcription output: L'API accepte les requêtes POST avec un corps JSON contenant l'identifiant de l'utilisateur et le jeton d'authentification. La réponse inclut le statut de la transaction et un code de confirmation.
After translation: The API accepts POST requests with a JSON body containing the user ID and authentication token. The response includes the transaction status and a confirmation code.
"Wir haben beschlossen, das Produkt erst im dritten Quartal zu veröffentlichen, um mehr Zeit für Qualitätssicherung zu haben. Das Marketing-Team soll bis Ende des Monats einen überarbeiteten Zeitplan vorlegen."
Transcription output: Wir haben beschlossen, das Produkt erst im dritten Quartal zu veröffentlichen, um mehr Zeit für Qualitätssicherung zu haben. Das Marketing-Team soll bis Ende des Monats einen überarbeiteten Zeitplan vorlegen.
After translation: We have decided to release the product in the third quarter to allow more time for quality assurance. The marketing team should submit a revised timeline by the end of the month.
Speak in complete thoughts. Whisper's language model works best with full sentences that provide context. Instead of dictating "meeting... Friday... postponed," say the complete sentence: "The Friday meeting has been postponed to next week." The surrounding context dramatically improves transcription accuracy, which in turn produces better translations. For more on speaking effectively with Steno, see our guide on writing in multiple languages.
Set the correct source language. Always match Steno's language setting to the language you are speaking. While Whisper can sometimes auto-detect languages, explicitly setting it eliminates ambiguity, especially for closely related languages like Spanish and Portuguese, or Norwegian and Swedish, where words overlap frequently.
Use DeepL for European languages. For translation between European languages, DeepL consistently produces more natural and accurate translations than other services. For Asian languages or less common language pairs, Google Translate often has broader coverage. Choose your translation tool based on the specific language pair you are working with.
Build a personal glossary. If you regularly translate content in a specific domain — legal, medical, technical — keep a reference document of key terms and their approved translations. After Steno transcribes your speech, you can check domain-specific terms against this glossary before sending the text to machine translation. This catches errors that even good translation services miss. For organizing reference materials, check our guide on brainstorming with voice.
Consider translating a 300-word business email from Spanish to English:
The savings are even larger for languages with non-Latin scripts. Typing Japanese, Chinese, Arabic, or Hindi requires input methods that are significantly slower than typing in English. Voice dictation in these languages is often 4 to 5 times faster than typing, making the voice-first translation workflow dramatically more efficient. Learn more about how Steno's audio pipeline handles multiple languages.
Steno is a transcription tool; it converts speech to text in the language you speak. For translation, you pair Steno's accurate transcription with a translation service like DeepL or Google Translate. This two-step approach gives you more control over both the source text accuracy and the translation quality than an all-in-one solution would.
Steno uses Groq Whisper, which supports over 50 languages including English, Spanish, French, German, Italian, Portuguese, Dutch, Polish, Russian, Japanese, Chinese (Mandarin), Korean, Hindi, Arabic, Turkish, Vietnamese, Thai, Indonesian, and many more. The full list is available in Steno's language settings.
Yes. Groq Whisper provides high accuracy across all supported languages, especially for clear speech in quiet environments. For professional translation work, the transcription step is actually more accurate than typing for many users, particularly when working in languages with complex scripts or unfamiliar keyboard layouts.
While Steno is not designed as a real-time interpreter, its fast transcription speed makes it useful for near-real-time translation in slower-paced conversations. Hold the hotkey, capture a few sentences, release, quickly translate the transcribed text, and respond. For fast-paced conversations, a dedicated interpretation tool would be more appropriate.
Speak naturally in any supported language and get accurate text for translation. Download Steno and simplify your multilingual workflow.
Download Steno for Mac