Coii VoiceInput recognizes speech in any language you speak into the microphone. If you write in two languages — English and Spanish, or Mandarin and Japanese — you can dictate in both without stopping to change a setting. The engine detects which language you are speaking, transcribes it correctly, and adapts the punctuation and spacing to match the language's rules.
Set up bilingual dictation
1. Add vocabulary for both languages
Open the app settings and add words to your dictionary that belong to each language. If you are an English speaker who writes in Spanish, add Spanish words that you use frequently: montaña, gracias, corazón. If you are a translator working between two languages, add terms from both.
The dictionary does not learn — it does exactly what you enter. Enter the words you want the engine to recognize without changing, and the engine uses those entries both before recognition (as hints) and after (as replacements).
2. Speak naturally
When you dictate, speak whichever language is natural in the moment. English one sentence, Spanish the next. Mid-sentence switches work too: "I spent the day en las montañas" transcribes exactly as spoken, with the language shift captured.
The engine detects the language transition and transcribes it. No picker, no button, no mode to switch — just speak.
3. Adjust text filters for each language if needed
Coii VoiceInput applies text filters in a fixed order: punctuation and spacing, then your dictionary, then optional filler removal. These filters understand multiple languages. Punctuation rules adapt automatically — an English sentence ends with a period and starts with a capital, and so does a Spanish one.
If you find that one language is being treated incorrectly (for example, if Spanish guillemets or other punctuation is being replaced), adjust your dictionary to include those characters and mark them as off-limits to the filters. But most of the time, the filters work without adjustment.
Real-world bilingual workflows
Translator working between English and Spanish: Set up a dictionary with industry terms in both languages. When you dictate, switch between languages as the source text requires. The engine keeps up with language boundaries and your translator-specific vocabulary stays consistent.
Spanish-language developer with English comments: Build code that is written in Spanish but documented in English. Dictate the Spanish logic, then switch to English for comments. The app transcribes both correctly within the same file.
Journalist reporting in two languages: Interview subjects who mix languages, or prepare pieces in both languages. Dictate the code-switched speech and let the engine handle the boundaries. Your dictionary keeps regional terms and names consistent.
Bilingual writer: Write articles or books where dialogue mixes languages, or where you are drawing on sources in both languages. Dictionary entries ensure that character names, place names, and repeated terms stay consistent across both languages.
Parent writing in two languages: Keeping a bilingual journal or writing letters to family in both languages. Dictate whichever language is closest to your thought at that moment. The app transcribes it without asking which mode you want.
How language detection works
The engine was trained on text in many languages, including the language-mixing patterns people actually use. When you speak English with Spanish words, it recognizes the boundary and transcribes both. When you speak two full sentences in different languages, the transition is captured.
Language detection is not perfect — every speech recognition system has failure modes — but it handles the most common patterns: a primary language with words and terms from another, code-switched sentences where bilingual speakers transition naturally between languages, and borrowed words that are standard in both languages.
When language detection struggles
Very rare language pairs: If you are mixing languages that are distant from each other and uncommon in the training data (for example, Icelandic and Tamil), detection may be less reliable. Test with a few sentences to see how your particular pair performs.
Words that exist in both languages: If you speak a word that exists identically in both your languages, the engine may guess the language incorrectly. This is where the dictionary helps — add the word as an entry so the engine knows what you meant.
Accents and regional differences: If you have an accent in one language, the engine adapts based on its training. If you have an accent in both, dictation might sometimes guess the wrong language. Testing and dictionary entries improve accuracy.
Very short utterances: A single word is harder to classify than a sentence. If you dictate one word, the engine has less context to decide which language you spoke. Full sentences are easier to recognize correctly.
Dictionary entries for bilingual writing
Your dictionary travels with your licence to every machine you activate it on. If you set up bilingual vocabulary on your Mac, and then activate your licence on a new Mac, the dictionary comes with it — your setup is portable. This is especially powerful for specialized vocabulary in fields like law, medicine, or technical work.
Entries work the same way in both languages: a dictionary entry is applied after recognition to fix words and apply your rewrites. If you have an entry btw → by the way, it applies whether you dictated in English or (if you speak both) when you used that abbreviation in an English phrase within Spanish text. Similarly, a term you teach the app will be recognized consistently across both languages.
Punctuation and spacing across languages
The text filters that handle punctuation and spacing understand multiple languages. Comma rules in English are different from Spanish — a thousand separator is a period in English and a comma in Spanish. The filters apply the correct conventions based on detecting what language each sentence is written in.
If you notice punctuation is being applied inconsistently across your two languages, adding a dictionary entry with the correct punctuation mark can override the automatic filtering. This is rarely necessary, but it is available if your particular language pair needs adjustment.
Accuracy and quality across both languages
The speech recognition engine runs the same whether you are dictating in one language or two. Accuracy depends on how clearly you speak, your microphone quality, and how well the recognition model handles the specific language pair you use. With two languages, the only new variable is whether the engine correctly identifies the language boundary — which it does with high accuracy for distinct language pairs.
If one of your languages is less common or you have a strong accent in one language, accuracy may vary. The best way to improve is to test with a few sentences and adjust your dictionary if certain terms are consistently misrecognized. The history feature lets you see what the raw transcript looked like before filtering, which helps diagnose whether the engine recognized the words correctly or the filters changed them.
If you have specific punctuation preferences that the filters are not applying correctly, you can add custom dictionary entries to override them. This is rare, but possible.
Started in English, switched to Spanish mid-paragraph: "I love dictating because es muy rápido para escribir." Both languages transcribed in one take.
Testing your bilingual setup
Start with one language: Make sure the app works well in your primary language before adding a second. This isolates any issues.
Add a few dictionary entries: Add the five or ten most important terms from the second language and test how they are recognized.
Dictate a mixed sentence: Once you are confident in each language separately, try speaking them together. Test a sentence that switches languages mid-stream.
Adjust the dictionary: If certain words are consistently misrecognized, add them to the dictionary and re-test.
Performance and accuracy in bilingual mode
Recognition speed and accuracy are the same whether you are dictating in one language or two. The engine processes whatever you speak, and the time to transcription is about half a second after you release the key — the same whether you used one language or both.
Accuracy depends on how distinct the languages are from each other and how much code-switching is happening. Distant language pairs (English and Mandarin) are easier to distinguish than similar ones (English and Spanish, or Norwegian and Swedish). Mixed-language sentences are recognized at the same accuracy as monolingual ones, because the training data includes bilingual speakers.
If you notice lower accuracy in bilingual mode, check whether your dictionary has the key terms entered. The dictionary helps the engine disambiguate when it is uncertain.
Switching to a new language
If you later want to add a third language, or switch from Spanish to French, update your dictionary entries. The app does not have a language setting — it recognizes whatever you speak — so changing your vocabulary is the only setup required.
Bilingual setup across your devices
If you use Coii VoiceInput on multiple Macs, your bilingual setup travels with you. Your dictionary is tied to your licence, so when you activate on a new machine, all of your vocabulary entries — in both languages — come with it. This means your first setup work pays off across every device you use.
Each machine stores its own history, so your past dictations don't sync between devices. But the vocabulary that matters for recognition is portable, making it easy to maintain consistency in your bilingual writing across machines without re-entering every term.
Why bilingual auto-detection matters
Most dictation software requires you to choose a language before you dictate, which breaks the flow if you work across languages. You are forced to pick English, dictate a sentence, then pick Spanish, dictate the next one. Over a long document that switches languages frequently, this mode-switching becomes a bottleneck.
Coii VoiceInput's approach — automatic language detection — lets you write the way bilingual people actually think and speak. You say whichever language is natural, and the app keeps up. This is especially valuable for code comments in mixed languages that borrow terms from both your languages, or any work where code-switching is the norm.
Compare to other approaches
Switching apps between languages: Some users keep a separate dictation app configured for each language. Coii VoiceInput does this in one tool — set it up once, then speak whichever language you need. No app switching, no mode picking.
Manually typing: For bilingual writers who are slower at typing, dictation becomes even more valuable when it handles both languages. The time savings compound, and the cognitive load of language mode-switching is eliminated.
Cloud-based options: Wispr Flow advertises support for 100+ languages, and you can select which language to use. Coii VoiceInput auto-detects instead, detecting language boundaries in real time. Both approaches work; this one requires no configuration mid-dictation.
FAQ
If I switch languages every sentence, will the app get confused?
No. The engine detects language per sentence. If you dictate in English, then Spanish, then English again, each sentence is transcribed in its language.
What if I have a very heavy accent in one language?
The engine adapts to accents based on its training data. Heavy accents may occasionally be misrecognized, but this is a recognition challenge that occurs whether you are bilingual or monolingual. Test your accent with a few dictations to see how well it performs.
Can I use the app in more than two languages?
Yes. The engine detects any language in its training data. Most users find one or two languages natural to their work, so the dictionary setup is focused there, but dictation works in more languages if you need it. The language auto-detection that handles two languages scales to three or more just as well.
Do I need to tell the app I am bilingual?
No. There is no "bilingual mode." The engine always detects the language you are speaking. Tell it your vocabulary by filling in the dictionary, and it handles the rest. The automatic detection means your dictation flow never interrupts for a language picker or mode switch — you just speak naturally and the app keeps up.
What about code-mixed speech where I'm switching every few words?
The engine handles this well within single sentences. If you are rapidly switching between languages — every few words rather than every sentence — accuracy may decrease slightly, but most bilingual speakers find that sentence-level or phrase-level code-switching is recognized reliably. Testing with your specific language pair is the best way to know how well it works for your pattern of mixing.