Languages and accents
Dictation Languages, Accents and Voice Translation
Every major platform now dictates dozens of languages, but support, accuracy and features thin out quickly beyond English and a dozen large languages, so the right tool depends on the exact language, regional variant and accent you speak. Set the precise language and region, use a multilingual model when you mix languages or speak with a strong accent, and treat translation as a second step whose errors add to the recognizer's.
Updated October 16, 202621 min read61 cited facts, each traced to its source
63language variants
Apple's macOS feature availability page lists 63 language and regional variants for Dictation on the Mac.
Apple, Apple, 2026
3.0percent word error rate
On the Fleurs benchmark, the Whisper large-v2 model had a word error rate of 3.0 percent for Spanish.
Alec Radford et al., OpenAI (arXiv), 2022
21.5percent word error rate
On the Fleurs benchmark, the Whisper large-v2 model had a word error rate of 21.5 percent for Hindi.
Alec Radford et al., OpenAI (arXiv), 2022
94.8percent word error rate
On the Fleurs benchmark, the Whisper large-v2 model had a word error rate of 94.8 percent for Yoruba.
Alec Radford et al., OpenAI (arXiv), 2022
Guides in this topic
Each guide takes one part of the subject in depth.
- Dictation Languages: What Each App and Device SupportsMajor dictation tools cover many languages, but no two vendors publish the same list, and their numbers count different things. A language can be on a list and still need a connection, lack punctuation commands or be much less accurate than English, so check your exact language and region on the vendor's list and then test it on your own speech.
- How to Change the Dictation Language on Any DeviceAdd the dictation language in the place your device keeps it, then switch with the key or menu the device gives you: Keyboard settings on a Mac, the keyboards you add on an iPhone or iPad, the input language on Windows, and the languages enabled in Gboard on Android.
- Hindi Voice Typing and Speech to TextHindi voice typing comes built into Android, iPhone, Windows and Google Docs, so you can start without buying an app. Hindi is recognized less accurately than English, and mixed Hindi and English is harder still, so read the output before you send it.
- Speech to Text With Translation on Phones and ComputersSpeech translation is two steps: a recognizer turns your speech into text, then a translator turns that text into another language. Phone translator apps, translating keyboards and dictate-then-translate on a computer each do this with different limits, and a word the recognizer gets wrong becomes a wrong translation.
- Speech Recognition and Accents: Accuracy and FixesSpeech recognition still makes more errors on accented English than on standard US English, and the gap is measurable. Research points largely to the speech a system was trained on. The fixes you control are the exact regional language setting, a multilingual model for mixed speech, custom words, and a headset microphone.
- Multilingual Dictation: Two Languages, Auto-Detect, MixingMost built-in dictation listens for one language at a time. You can switch by hand with a tap or shortcut once both languages are set up, and automatic detection usually picks one language for a whole recording or request. A sentence that mixes two languages is the hardest case, and none of the vendor documents we read says its dictation follows a switch inside a sentence.
- Spanish Speech to Text and Talk to TextYou can dictate Spanish on Mac, Windows and in Word, and Microsoft lists both Spanish (Mexico) and Spanish (Spain) for Windows and Word. Recognizers handle Spanish well, but they can drop written accents such as the tilde, so check accent marks in every draft. To get English, dictate the Spanish first and translate it as a separate step.
- Japanese, Chinese and Korean Speech to TextDictation in Japanese, Chinese and Korean writes kanji, hanzi or hangul straight from your speech, so the errors to watch for are homophones and spacing. Published accuracy for Japanese and Chinese is reported as character error rate, not word error rate, and Apple's Mac Dictation lists all three languages.
- How to Translate Audio and Video to English TextYou can get English text from a recording in another language in two ways: a model that translates the speech straight into English, or a transcript in the original language that you translate yourself. Whisper does the first step for free on your own computer, and translation quality falls sharply for low-resource languages, so check any translation that matters.
- Dictation for Language Learning: How and Why It WorksDictation is a listening exercise: you hear a short passage, write it down and compare your version with the original. The strongest study we found, a 2002 controlled trial in Tehran, showed a significant gain in listening after frequent dictation, but it is the only controlled trial we found that measured listening.
- Whisper Supported Languages and Accuracy by LanguageWhisper supports 98 or 99 languages, depending on the source, and its accuracy follows how much training audio each language had. In the FLEURS benchmark, the large-v2 model had a word error rate of 3.0 percent in Spanish and 94.8 percent in Yoruba, so check your language against test numbers, not the list.
- Arabic Speech to Text and Right-to-Left DictationBuilt-in Arabic dictation covers a few country variants on Apple devices and in Microsoft Word, but Windows 11 voice typing does not list Arabic. Accuracy falls most on dialects, and right-to-left text can show numbers and punctuation in an unexpected order.
- Speech to Text in French, German, Russian and MoreDictation covers French, German, Portuguese, Italian, Dutch and Russian on the Mac and in Windows 11, but the regional variants differ from one platform to the next. On a read-speech benchmark, Whisper large-v2 has an error rate of 8.3 percent for French and 4.5 percent for German. On a Swiss German test set, Whisper large-v3, used without training on Swiss German, had an error rate of 23.0 percent.
- Dictation Punctuation Commands in Other LanguagesEach language has its own spoken names for punctuation and its own typographic rules, so an English command can give the wrong mark. Spanish needs opening ¿ and ¡ marks, French puts a non-breaking space before the colon, German opens quotations with a low mark, and Chinese uses full-width marks, while Japanese dictation commands give half-width ones.
You can dictate in dozens of languages for free on most recent phones and computers, but how well it works depends on your language far more than on the app. Apple lists 63 language and regional variants for Dictation on the Mac 1, and Google Cloud advertises over 125 languages for its Speech-to-Text service 2. Behind those headline counts, accuracy follows how much recorded speech a language has been trained on 3. On a read-speech benchmark, the same Whisper model has a word error rate of 3.0 percent in Spanish 4, 21.5 percent in Hindi 5 and 94.8 percent in Yoruba 6, which is close to unusable.
Four rules cover most of what goes wrong. First, choose the exact language and region you speak, not just the language. Second, check that the language works in the mode you want (on the device or online, with automatic punctuation or without). Third, if you switch between languages, or mix them in one sentence, use a tool that was built for it, because most built-in dictation listens for one language at a time. Fourth, treat voice translation as two steps, recognition then translation, because the mistakes of the first step become mistakes in the second.
This page is the overview of the whole subject. Each section gives the short answer and the evidence, then points to the guide that goes deeper and to the questions people ask most. If you are new to talking instead of typing, start with our overview of dictation, and for how the recognizers themselves work see speech recognition.
Every major platform dictates dozens of languages, but no two lists match
Most of the big platforms cover a shared core: English, Spanish, French, German, Japanese and Chinese. Beyond that core the lists diverge, and the numbers are not comparable, because each vendor counts something different (languages, regional variants, or languages per feature). The table puts the checked figures side by side.
| Tool | What the vendor lists | Runs on the device? | Notes |
|---|---|---|---|
| Apple Dictation on Mac | 63 language and regional variants 1 | 48 variants work on the device after a model download 7 | Automatic punctuation in 33 variants 8 |
| Apple Dictation on iPhone | Not available in all languages or regions 9 | Processed on the iPhone in many languages, no internet needed 10 | Language follows the keyboards you add 11 |
| Windows 11 voice typing | A list of languages by region, including Hindi, Japanese and Korean 12 | No: online recognition powered by Azure 13 | Follows the input language 14 |
| Windows Voice Access | 15 languages and dialects 15 | Not checked | A voice control tool. Check Microsoft's setup page, since its FAQ gives a shorter list |
| Google Cloud Speech-to-Text | Over 125 languages 2 | No, a cloud service | For developers |
| Whisper (OpenAI, open source) | 99 languages 16 | Yes, on your own computer | 680,000 hours of training audio 17 |
| Meta MMS (research) | 1,107 languages in one model 18 | Research release | Not a finished dictation app |
Three points follow from the table. A longer list is not a better tool: Apple's Mac list is long, but only about three quarters of its variants also appear on the on-device list 19. Lists change with each release, so the vendor's own page beats any article, including this one. And the headline counts hide a gap that matters more than the count, which is how accurate each language is (covered below).
Our guide to dictation languages by app and device compares Apple, Microsoft, Google, Dragon and Whisper in one table, with every row linked to the vendor's own list. It also answers which dictation app supports the most languages, what languages Dragon NaturallySpeaking supports and how many languages Google Speech-to-Text supports.
Changing the dictation language is a setting you must find on each device
The dictation language is usually its own setting, separate from the language your screen is shown in. That is why dictation can fail on a computer set up in English when you start speaking French: the recognizer is still listening for English. Apple's checklist for Mac dictation problems tells you to choose the correct language and region 20.
The steps differ by device, and so does the rule that decides which language is used:
- Mac. In Keyboard settings, open Dictation, click the Edit button next to Languages and select a language and region 21. With more than one language set up, you switch by clicking the language label next to the cursor or by pressing the Globe key and choosing a language 22.
- iPhone and iPad. Dictation works in a language when you use a keyboard for that language 11. See the phones and tablets overview.
- Windows. Voice typing follows the input language you have selected 14. See the Windows overview.
- Android and Chromebook. Changing the language in Gboard does not change the language of the Android device 23. On a Chromebook, dictation does not work if the language you speak does not match the device language 24.
- Google Docs. Voice typing works in the latest versions of Chrome, Edge and Safari, and the browser controls the speech service 25.
The guide on how to change the dictation language on any device gives the exact menus for each platform and the OS version they describe. For specific problems, read why dictation says it is not available in the current language, how to switch between dictation languages quickly, how to dictate in a language your device does not support and how to change the dictation language on Apple TV. Mac users can also read the page on fixing the Mac dictation language.
Accuracy follows the hours of training audio, so Spanish beats Hindi and Hindi beats Yoruba
Support is a yes or no. Accuracy is a number, and for a single model it ranges from 3.0 percent in Spanish 4 to 94.8 percent in Yoruba 6, across the languages it claims to cover. The clearest published evidence is the Whisper paper, which reports a word error rate for each language on the Fleurs read-speech benchmark. Of Whisper's 680,000 training hours 17, only 117,000 hours cover languages other than English (96 of them) 26, and the result tracks that supply of data closely.
Spanish scores 3.0 percent 4, German 4.5 27, Russian 5.6 28, French 8.3 29, Arabic 16.0 30, Hindi 21.5 5, Swahili 39.3 31 and Yoruba 94.8 6. The authors fit the pattern: across languages, the log of the error rate has a squared correlation of 0.83 with the log of the training data 3, and they estimate that the error rate halves for every 16 times more data 32.
Read the chart as a guide to relative difficulty, not as a promise. These are the scores of one model on read speech in good audio. Your own dictation, with your microphone and your vocabulary, will score differently, and other recognizers rank the languages differently. Meta's authors make the same point from the other side: current speech technology is restricted to about one hundred languages, a small fraction of the more than 7,000 spoken in the world 33. A low-resource language is one with too little recorded and transcribed speech to train on, and for those the honest answer is often "try it on your own voice first."
Our guide to Whisper's languages and accuracy groups the languages by measured error rate and explains how Whisper detects, or misdetects, the language. Related questions: can Whisper translate audio, how good is Whisper at Hindi and why Whisper transcribes in the wrong language. For the general question, read is speech recognition less accurate in languages other than English. The wider research on error rates sits in the overview of speech recognition.
Accents and dialects lower accuracy, mostly because of the training data
An accent is a way of pronouncing a language, and a dialect also differs in words and grammar. Recognizers do worse on speech that differs from most of their training audio, and that has been measured. In a 2017 study of YouTube's automatic captions, accuracy was lower for women and for speakers from Scotland 34. In a 2020 study of five commercial systems, the average word error rate was 0.35 for Black speakers against 0.19 for white speakers 35, and the authors traced the gap to the acoustic models 36.
The evidence is not one-sided. A 2004 study of one speech corpus found no evidence that accent or speaker age affected recognition performance 37. The practical reading: accent is one factor among several, and a good microphone and steady pace can matter as much as how a speaker sounds.
What helps, in order of effect:
- Pick the closest regional setting. Some tools list separate English options, such as English (India) and English (Australia) 15.
- Use a large multilingual model such as Whisper. Its authors report that, compared with humans, the models approach human accuracy and robustness 38.
- Add custom vocabulary for names and terms, and say them the same way each time.
- Speak a little slower and closer to the microphone, with steady volume, rather than exaggerating pronunciation.
- Test before you buy. Read a paragraph you wrote yourself, and compare the output to the text.
The guide to speech recognition and accents quantifies the gap from the studies and gives the plan. The questions are which speech to text works best with an Indian accent, why voice recognition does not understand my accent, whether you can train dictation to your accent, how non-native speakers can get accurate English voice typing and whether to set dictation to US or UK English. Definitions: accent, non-native speech and locale. The research on bias in recognizers is covered in the speech recognition overview, and the access implications in the accessibility overview.
Most dictation listens for one language at a time, and mixing languages is the hardest case
If you write in two languages, the question is how the tool knows which one you are speaking. There are three designs. A tool can listen for one language you choose, which is how most built-in dictation works: you switch by hand, with a tap on a label or a keyboard shortcut. A tool can detect the language automatically from the audio, as Whisper does. Or a tool can follow a single sentence that mixes two languages, which is called code-switching.
On Apple's Mac, for example, switching means clicking the language label or pressing the Globe key once more than one language is set up 22. That is quick, but it is still a switch, and a name or a borrowed word in the other language can be misheard or turned into a word in the active one. Automatic spoken language identification removes the manual step, but it chooses one language for a stretch of audio, so a recording that changes language halfway can come out half wrong. Recognition of code-switched speech is still an active research problem rather than a solved feature, and multilingual models such as Whisper often struggle with it 39.
A dependable routine for people who write in two languages: set the language by hand for each session, say foreign names slowly or spell them, add the names you use often to a vocabulary list, and dictate the sentence in one language, then edit. The guide to multilingual dictation gives a daily workflow and the checked claims about each tool. The questions: can you dictate in two languages at the same time, does dictation detect the language automatically, can speech to text handle Hinglish, Spanglish and other mixed languages and how to dictate foreign names and words correctly.
Spanish, French, German and the other big European languages are the best supported
For the large European languages, dictation is mature on every platform. Whisper's error rates on read speech are 3.0 percent in Spanish 4, 4.5 percent in German 27, 5.6 percent in Russian 28 and 8.3 percent in French 29, compared with 4.2 percent in English 40. Windows 11 voice typing lists two Spanish variants, Spanish (Mexico) and Spanish (Spain) 41, so choose the one closest to your speech.
What goes wrong is specific to each language's grammar and writing. In French, homophones and agreement endings are the usual slips, because many different spellings sound the same. In German, long compound words and the capitalization of every noun test the recognizer. In Portuguese, the Brazilian and European varieties differ enough that choosing the wrong one costs accuracy. In Spanish, you do not need to type the accent marks or the ñ, because the recognizer supplies them. Italian and Dutch are well supported with fewer regional options, and Russian is good, with case endings worth checking. Smaller European languages are patchier, especially for offline use.
Punctuation is where these languages differ most in practice, and it is spoken in the language you are dictating. Spanish question marks are double signs, an opening mark and a closing mark 42, so a good Spanish dictation tool must insert both. French puts a space before colons and question marks and uses guillemets for quotations, and German opens quotations with a low mark. Automatic punctuation is available in fewer languages than dictation itself: Apple, for example, lists it for 33 of its 63 Mac variants 8, about 52 percent 43. In Microsoft 365, the automatic punctuation setting works only where it is available for the language you chose 44.
The guides in depth: Spanish speech to text, speech to text in French, German, Russian and more and dictation punctuation commands in other languages. Questions: talk to text in Spanish and get English, does Spanish dictation add accent marks, which Spanish to choose, is speech to text as accurate in Spanish as in English, how to voice type in Russian, Brazilian and European Portuguese, Swiss German, and the punctuation questions for Spanish, French and German. Our general guide to dictation punctuation in English is in the dictation overview.
Japanese, Chinese and Korean dictation skips the slowest part of typing
In these languages typing means building characters through an input method editor: you type sounds, often as pinyin in Chinese, and then pick the right characters from a list. Speech skips that conversion and goes straight to characters. In a controlled phone study, Mandarin speech input reached 123 words per minute 45 against 43 words per minute on the built-in Pinyin keyboard 46, close to the English ratio of 153 against 52 47, 48.
The errors are different from those in alphabet languages. Because many characters share a sound, the main error is a homophone: the right sound written as the wrong character. Accuracy is therefore measured in characters, as a character error rate, not in words, since these scripts do not put spaces between words. Whisper's character error rate on Japanese is 5.3 percent on the Fleurs benchmark 49. Windows 11 voice typing lists Japanese and Korean 12. On the Mac, Apple's automatic punctuation list names Cantonese and Chinese in separate entries, such as Chinese (Simplified) and Chinese (Traditional) 50, so check which variant you need. Punctuation by voice produces full-width marks such as the ideographic full stop, and Korean dictation has its own slips with word spacing and loanwords.
The guide to Japanese, Chinese and Korean speech to text compares the options with published error rates. Questions: the best speech to text for Japanese, how to voice type in Chinese, how to voice type in Korean, is speaking Chinese faster than typing pinyin and how fast people type in Japanese. For the typing side, see our overview of typing speed.
Hindi and the Indian languages are supported, but less accurately than English
Hindi voice typing is available in Google Docs 51, and Windows 11 lists Hindi and Marathi among its voice typing languages 12. Accuracy is the weak point: Whisper's error rate on Hindi is 21.5 percent 5, and Telugu is far worse at 99.0 52. Tamil does better at 17.5 53, Urdu at 22.6 54.
Everyday Hindi adds two problems. Many people mix Hindi and English in one sentence (Hinglish), which is code-switching, and multilingual models often struggle with it 39. And Hindi dictation needs the purna viram (the full stop, danda) and other punctuation said in Hindi. For Hindi speakers who want English text, speaking Hindi and getting English output is a translation task, discussed below.
The guide to Hindi voice typing and speech to text compares every free option and extends to Bengali, Tamil, Telugu, Marathi, Gujarati and Urdu. Questions: Hindi voice typing on a computer, Hindi voice typing to English, which speech to text API works best for Indian languages, Urdu voice typing and a good Hindi typing speed. Related: transliteration is writing one language in another's script, which is how Hinglish is often typed.
Arabic and other right-to-left languages add dialects and display problems
Arabic raises two separate difficulties. The first is dialect. Most dictation targets Modern Standard Arabic, the formal written language, plus a few country variants, while people speak regional dialects that differ in vocabulary and sound. Dialects also have no standard spelling, so even deciding what the right transcript is can be a problem. Whisper's error rate on Arabic is 16.0 percent 30, between Russian and Hindi in the chart above. Dictated Arabic also usually leaves out the short-vowel marks ( diacritics) that formal text sometimes carries, which suits everyday writing but not teaching texts.
The second is display. A line of Arabic or Hebrew that contains digits or an English word is bidirectional text: the Unicode standard says that when right-to-left scripts use digits written from left to right, the text is a mixture of both directions 55. Dictation that inserts a number or a Latin word into a right-to-left sentence can therefore look out of order, depending on the editor. Setting the paragraph direction in the editor often fixes the display.
Support is uneven across the right-to-left languages. Windows 11 voice typing lists Hindi, Japanese and Korean 12, but when we checked on October 16, 2026 the list did not name Arabic or Hebrew. Urdu scores 22.6 on Whisper 54, and Persian 32.9 56. Read the right-to-left script entry for the terms.
The guide to Arabic speech to text and right-to-left dictation covers platforms, dialects and editors. Questions: does speech to text understand Arabic dialects, the best Arabic speech to text API, why dictated Arabic or Hebrew text looks backwards and how to learn touch typing in Arabic.
Speaking in one language and getting text in another is two steps, and the errors add up
Voice translation chains two systems: a recognizer writes down what you said, and a translator converts that text. A mistake in the first step is a mistake in the second, because the translator never hears the audio. Meta's authors write that conventional speech-to-speech translation systems rely on cascaded systems that perform translation progressively 57, and their newer SeamlessM4T model aims to do recognition and translation in a single model for up to 100 languages 58.
There are three practical ways to do it:
- Translator apps (Google Translate, Microsoft Translator, Apple Translate) are built for conversations: you speak, they show and read back the translation.
- Translating keyboards put translated text straight into any app.
- Dictate, then translate on a computer: dictate in your own language where recognition is best, then translate the text. You can read the original and the translation side by side, which catches errors.
Live translation works best with short, clear sentences and common language pairs. Privacy differs by tool: voice translation stays on your device only if both steps run on it, and many translators send audio to a server. For anything confidential, check where each step runs. Our guide on speech to text with translation compares the three routes and explains the research on accuracy. Questions: using voice input in Google Translate, why Google Translate voice is not working, how accurate speech to text translation is and which APIs combine speech to text and translation. The definition is speech translation.
Whisper can translate a recording into English, and only into English
For a recording rather than live speech, the working methods are Whisper's translate mode, transcribing then translating, and phone features. Whisper translates speech from its supported languages into English only, and the repository warns that the turbo model is not trained for translation, so for translation pick a multilingual model such as medium or large 59. If you need to check the original wording, transcribe first and translate second: you keep a transcript in the source language to compare against.
Quality follows the same data curve as recognition. A language with little training audio transcribes poorly, and what is translated from a poor transcript is worse. Yoruba and Swahili sit at the weak end (Swahili at 39.3 31 and Yoruba at 94.8 6 word error rate on read speech), so check any important passage with a native speaker. Confidential recordings are best kept on your own device, using a local model. A hosted service sends the audio to a third party, which matters for legal, medical or business material, as covered in our overview of transcription.
The guide to translating audio and video to English text gives each method with its cost and accuracy. Questions: translate Japanese audio to English, translate Zulu or Xhosa audio to English, translate a voice note to text and translate a video into English subtitles. For free file transcription in a browser, see audio to text, and for dictating online see online dictation.
Languages differ in how fast they are spoken, but not in how much they say
If you are comparing dictation speed across languages, words per minute is a poor yardstick, because words differ in length. Researchers studying 17 languages found that languages encode a similar rate of information, about 39 bits per second, despite wide differences in speech rate and in information per syllable 60. For dictation, this means a language that sounds slower is not less informative per second, so it is not a worse language to dictate in. The effect of speaking on typing is what matters, and our overview of typing speed covers how speech compares with the keyboard.
The question do people speak faster in some languages is answered in full in the guide to dictation languages.
Dictation also works as a way to learn a language
Dictation is an old language-teaching exercise: a teacher reads a passage and students write it down. Cambridge's dictionary defines it in schools as a test in which a teacher reads a passage aloud and students learning a language write it down 61. It trains listening, spelling and grammar in a single task. The method is to listen, write, compare and repeat, with audio slightly above your level, and variations such as partial dictation and dictogloss add variety.
Speech recognition lets you turn it around. Instead of writing what you hear, you speak and read what the recognizer wrote: if it understands you, your pronunciation is clear enough for the words it chose, and if not, you see where. It is a coarse check, because a recognizer is more forgiving than a person in some ways and less in others. The guide to dictation for language learning explains the exercise, the evidence and how to use voice typing for pronunciation. Questions: is dictation good for language learning, how to practice dictation in English, words for an English dictation test for kids, Hindi dictation words without matra, what dictation is called in Hindi and using voice typing to practice pronunciation.
How to choose a tool for your language
Start from the language, not the app, and test on your own voice. A short procedure:
- Find your exact language and region on the vendor's list: Apple's, Microsoft's or Google's own pages, not a roundup.
- Check the mode you need: on the device or online, automatic punctuation, voice commands. Many features arrive in English first.
- Dictate a paragraph of your real writing, with your names and terms, and count the errors. Anything under a few errors per hundred words is comfortable to edit.
- If accuracy is poor, change the region, move closer to the microphone, add vocabulary, or try a multilingual model.
- If you mix languages, choose a tool built for it, or dictate in one language at a time.
The built-in free tools are enough for most people in the big languages: Apple Dictation on the Mac and iPhone, Windows voice typing, and Gboard on Android. A dedicated app is worth considering when you need a multilingual model, on-device privacy or custom vocabulary. Yap, which we make, is an on-device dictation app for Mac and Windows, and its download page and pricing page list the languages and plans it supports today. For platform specifics, read our overviews for the Mac, Windows and phones and tablets, and for choosing among apps see how to choose a dictation app.
Frequently asked questions
How many languages can dictation handle? It depends on the tool: Apple lists 63 variants for the Mac 1, Google Cloud over 125 languages 2, and Meta's research model 1,107 18. Support is not accuracy, which falls steeply for languages with little training data.
Is dictation less accurate with an accent? Often, yes, but by an amount that varies. The gap comes from the training data, not from how well you speak, and the right regional setting, a multilingual model and a good microphone narrow it.
Can I dictate in two languages? Yes, by switching language by hand, and with a multilingual model in some tools. Mixing two languages inside one sentence is the hardest case.
Can dictation translate as I speak? Translator apps do, in two steps, and each step adds errors. For recordings, Whisper translates into English only.
Where to go next
The guides in this subject, in the order most readers need them:
- Dictation languages by app and device
- How to change the dictation language
- Hindi voice typing
- Speech to text with translation
- Speech recognition and accents
- Multilingual dictation
- Spanish speech to text
- Japanese, Chinese and Korean speech to text
- Translate audio and video to English
- Dictation for language learning
- Whisper languages and accuracy
- Arabic speech to text and right-to-left dictation
- Speech to text in French, German, Russian and more
- Dictation punctuation commands in other languages
For the neighboring subjects, see dictation, speech recognition, transcription, the Mac, Windows and phones and tablets overviews, accessibility and typing speed.
Evidence
Every number on this page, with where it comes from. Open a fact to see the exact words in its source.
- 1.
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
- 8.
- 9.
- 10.
- 11.
- 12.
- 13.
- 14.
- 15.
- 16.
- 17.
- 18.
- 19.About 76 percent of the Mac Dictation language variants also appear on the on-device list.
Calculated from 2 other facts
- 20.
- 21.
- 22.
- 23.
- 24.
- 25.
- 26.
- 27.
- 28.
- 29.
- 30.
- 31.
- 32.
- 33.
- 34.
- 35.
- 36.
- 37.
- 38.
- 39.
- 40.
- 41.
- 42.
- 43.About 52 percent of the Mac Dictation language variants have automatic punctuation.
Calculated from 2 other facts
- 44.
- 45.
- 46.
- 47.
- 48.
- 49.
- 50.
- 51.
- 52.
- 53.
- 54.
- 55.
- 56.
- 57.
- 58.
- 59.
- 60.
- 61.
Questions people ask
- Can speech to text handle Hinglish, Spanglish and other mixed languages?Only partly, and the error rates are high. A 2025 study of Mandarin-English and Spanish-English speech found Whisper's mixed error rate on code-switched speech was about 39 percent, and the vendor documents we read do not say their detection follows a switch inside a sentence. A 2023 study reports 14.2 percent for a Whisper model adapted to Mandarin-English mixing, in a research setup, so test a tool on your own sentences before you rely on it.
- Can Whisper translate audio?Yes. Whisper can translate speech from other languages into English, but its translate task produces English only. The turbo model is not trained for translation, so use a multilingual model such as medium or large for that step.
- Can you dictate in two languages at the same time?Not inside one sentence, as far as the vendor documents we read show. Most built-in dictation works in one language at a time, so you set up both languages and switch between them by hand, and some tools choose a language for a whole request. None of the documents we read says a switch can happen in the middle of a sentence.
- Can you train dictation software to your accent?Usually not in the way the word suggests. The dictation tools we checked do not let you retrain their core model, but some learn a voice profile or a custom vocabulary. Research shows that adapting a model to a speaker can cut errors on accented speech a great deal, so check what your app offers, then measure the change with the same passage before and after.
- Can you translate Zulu or Xhosa audio to English?Yes, but expect weaker results than for widely spoken languages. Transcribe the audio in Zulu or Xhosa with a tool that supports that language, check the transcript against the recording, then translate the text. Whisper's translate task is a weak choice for Xhosa, because its training data has no isiXhosa.
- Can you use voice typing to practice pronunciation?Partly. Voice typing can show you where the recognizer heard something other than what you meant, which is a useful clue. General voice typing is built to look past mispronunciations, so a trainer that flags errors is the better tool for pronunciation.
- Do people speak faster in some languages?Yes, in one controlled study Japanese was spoken faster than English, measured in syllables per second. But we found no reliable words-per-minute figure for each language, because words are counted differently in each one, and speaking rates vary widely from one conversation to the next. Measure your own pace in one language, and do not rank languages by words per minute.
- Does dictation detect the language automatically?Some do, but usually for a whole request or recording, not for each word. Gboard's advanced voice typing detects your spoken language in six supported languages, Google and Azure choose one language from a list you supply, and Whisper detects it from up to the first 30 seconds of audio. Apple's and Microsoft's voice typing pages we read describe choosing the language yourself.
- Does dictation work for both Brazilian and European Portuguese?Yes. Apple's Dictation for the Mac and Windows 11 voice typing both list Brazilian and European Portuguese as separate choices, so you pick the variant you write in. The exception is the Mac's on-device list, which includes Portuguese (Brazil) but not Portuguese (Portugal). We found no published study that compares the two variants on the same system.
- Does Spanish dictation add accent marks automatically?Usually, but not reliably. The recognizer chooses each accent mark from how the word sounds, and it can drop a tilde. The Spanish command list for Dictate in Word we checked includes opening marks but no accent-mark command, so read every draft for missing accents, especially on words that change meaning.
- Does speech recognition understand Swiss German?Partly. No major platform lists Swiss German as its own language. Apple lists German (Switzerland) and Windows 11 lists German with no region. In a 2024 test, Whisper large-v3, used without training on Swiss German, had a 23.0 percent word error rate, and the authors conclude it can transcribe Swiss German into Standard German without adaptation.
- Does speech to text understand Arabic dialects?Partly. Speech to text tools handle Modern Standard Arabic best and make many more errors on everyday dialects, and the size of the gap depends on the dialect, the tool and the recording. Test your own speech in the variety you use before you rely on any tool.
- Does Whisper support Arabic?Yes. Whisper's language list includes Arabic, and the large-v2 model had a word error rate of 16.0 percent on the FLEURS read-speech test. Accuracy drops on dialects, and on the Common Voice 9 test of volunteer recordings the same model had a word error rate of 53.8 percent, so test Whisper on the variety you speak before you rely on it.
- How accurate is speech to text translation?No single figure describes speech translation accuracy, because it has two stages and each depends on the language. The recognizer mishears words, and the translator then translates those wrong words faithfully, so recognition errors are the first thing to check. Test your own language pair on your own speech before you rely on it.
- How can non-native speakers get accurate English voice typing?Choose the English variant you speak and write, use a headset close to your mouth, speak in steady phrases, and test each tool on a passage you know. Research on non-native English finds the best systems are accurate on read sentences, while free speech is harder and each system handles filler words differently, so check the names, numbers and words that matter before you send.
- How can you dictate in a language your device does not support?You generally cannot make built-in dictation work in a language the device does not support, because a keyboard or setting does not create support that is missing. Your realistic options are a platform or service whose list includes your language, a dedicated app that lists it, or dictating in a language you can use and translating the text afterwards. Check the exact language and region in each list before you commit.
- How can you type faster in Hindi?Pick one Hindi input method and practise it every day, measuring your net words per minute so that correct words are what you count. Transliteration lets you type Hindi with the English letters you already know, and InScript is the standard layout for Indian scripts. No study we found compares the two for speed, so test both on your own text before you commit.
- How can you type Japanese faster?No study we found shows that romaji or kana input is faster for Japanese, so test your own input method before you change it, and consider speech for long text. Touch typing did not make typists faster in one 2016 study, and word prediction did not improve accuracy in a 2024 study of blind smartphone users.
- How do you add a dictation language on a Mac?To add a dictation language on a Mac, open System Settings, click Keyboard, then under Dictation click Edit next to Languages and select a language and region. Dictation must be turned on first, and once you have added more than one language you switch between them with the language label next to the cursor or the Globe key.
- How do you change the dictation language on Apple TV?On Apple TV, Apple documents one language setting for Siri: go to Settings, then General, then Siri, then Language. Apple's pages do not name a separate dictation language, so check the Siri language first. You turn dictation itself on or off under General, then Dictation.
- How do you change the voice typing language on Android?On Android, voice typing runs through Gboard, so you change the voice typing language by adding it to Gboard first, under Settings, then Languages, then Add keyboard. Then open voice typing, tap the abbreviation at the bottom left of the toolbar and select the language. On Pixel 6 or later, advanced voice typing can detect the language for you, but only for six languages and only when your Gboard and device languages match.
- How do you dictate foreign names and words correctly?Set the language first, then add each name or foreign word to the tool's custom vocabulary with its exact spelling, including accents, where the tool offers a list. Where there is no list, spell the word out or type it after dictation, and proofread the result. On a Mac, Voice Control accepts up to 1,000 vocabulary terms, but its Vocabulary option is dimmed for languages that do not support it.
- How do you dictate in French?To dictate in French, add a French variant in Dictation on a Mac or in Windows 11 Language & region, then start dictation in a text field. Choose French (France) or French (Canada) on Windows, and one of four French variants on the Mac, because the variants differ. Check the words that sound the same, such as et and est, and the verb and plural endings, before you rely on the text.
- How do you dictate in German?To dictate in German, add German to your dictation languages, then start dictation in the text field where you want the words. On a Mac, German is set under Dictation in Keyboard settings. In Windows 11, German is one voice typing language with no region, and it needs an internet connection, so check compound words and noun capitals before you rely on the text.
- How do you dictate in Japanese on a phone or computer?Turn on Japanese in your device's dictation or voice typing settings, click into a text field, and speak in Japanese. Check each sentence for homophones before you send it. Windows voice typing needs an internet connection, so use an offline option if you are not online.
- How do you dictate punctuation in French?Say the French name of the mark: point d'interrogation gives ? on the Mac, and deux points gives the colon in Microsoft Word. French quotations use guillemets, « and », which Apple's French table names guillemet ouvrant and guillemet fermant. The Québec language office's spacing table puts a non-breaking space before the colon, which a dictation tool may not insert, so check each colon.
- How do you dictate punctuation in German?Say the German name of each mark: Fragezeichen gives ? and Komma gives the comma on a Mac. German quotations open with a low mark („) and close with a high one (“), as Apple's German table gives, but Microsoft's German table maps the opening quotation phrase to guillemets (« »), so check the first quotation in each document.
- How do you do Hindi voice typing on a computer?On a Windows 11 PC, use the built-in voice typing, which includes Hindi. In a browser, Google Docs voice typing had Hindi on its list in Chrome in 2020, so check the current list there. On a Mac, check Dictation for Hindi in your own settings, because we could not confirm it from a page we could read.
- How do you do voice typing in Hindi on Android?On Android phones, add the Hindi keyboard in Gboard, tap the microphone at the top of the keyboard, and speak Hindi when you see "Speak now". Gboard has accepted Hindi voice input since at least August 2017, but Hindi is recognized less accurately than English, so read the text before you send it.
- How do you learn touch typing in Arabic?Learn Arabic touch typing the same way as English: use the Arabic layout your computer already has, learn the home-row letters first, and practice in short daily sessions with your eyes on the text. A 2024 study of 30 students reported that learners reached proficiency within three sessions, but it was small and had no control group.
- How do you learn touch typing in Hindi?To learn touch typing in Hindi, learn the InScript layout key by key, starting from the home row, and practise in short daily sessions while you keep your eyes off the keyboard. Expect it to take many hours of practice. If you already touch type on an English keyboard, you can type Hindi with transliteration without relearning the keys.
- How do you learn touch typing in Russian?Learn touch typing on the standard Russian layout, ЙЦУКЕН (JCUKEN), using the same finger rules you would use for English: each finger has a home position and returns to it after each key. Add Russian to your system, practise with a trainer that uses ЙЦУКЕН, and keep your eyes on the screen. We found no study that measures touch typing on a Cyrillic layout, so the research below comes from English typing.
- How do you practice dictation in English?Practice English dictation in short sessions most days. Write one or two sentences from audio that has a transcript, compare your version with the original, and dictate the same passage again a day or two later. Choose English audio slightly above your level, and set your device to the English variant the speaker uses.
- How do you say punctuation in Spanish dictation?Say the Spanish name of the mark: signo de interrogación gives the closing question mark on a Mac. Spanish questions and exclamations also need an opening ¿ or ¡, which Apple's Spanish Mac command list does not include, so type the opening mark yourself, or say signo de apertura de interrogación in Microsoft Word.
- How do you switch between dictation languages quickly?The quickest way is the key or control your device already provides: press the Globe key on a Mac, press Windows logo key + Spacebar on Windows, or tap the language label on the Gboard toolbar on Android. Apple's Globe key steps for iPhone are on a page for the iPod touch, so test them on your phone. Add each language first, because a language that is not set up cannot be switched to. Apple describes switching between languages you have set up, not recognizing two languages in one sentence.
- How do you talk to text in Spanish and get English?Talk to text writes the language you speak, so it does not turn Spanish speech into English by itself. Dictate in Spanish, check the Spanish, and then translate it, or send a finished recording to a Whisper translation model that outputs English.
- How do you translate a video into English subtitles?Run Whisper's translate task on the video with the medium or large model and the SRT output format. It writes a numbered, timed English subtitle file, which you then check against the video, line by line, before you publish it.
- How do you translate a voice note to text?Transcribe the voice note in the language the speaker used, check the transcript against the recording, then translate the text. Phone recorders can do the first step, and some can translate the finished memo, but translating a wrong transcript carries its mistakes into the English.
- How do you translate an audio file into another language?Transcribe the audio in the language it was spoken, check the transcript against the recording, then translate the text into your target language. Some speech tools translate the sound directly, but many produce only English, and the direct tools cover a fixed list of languages, so confirm your target is on that list first.
- How do you translate Japanese audio to English?Run Whisper's translate task on the recording, name Japanese as the spoken language, and use the medium or large model, which writes English text on your own computer. On one published test, Whisper translated Japanese into English with a lower score than German, so read the English against the audio before you rely on it.
- How do you use voice input in Google Translate?Open Google Translate, choose the language you speak and the translation language, then tap Speak on Android or Dictation on an iPhone or iPad and talk when it prompts you. On a computer, use Chrome, allow the microphone in the browser, and choose the spoken language yourself, because Translate does not detect it.
- How do you voice type in Arabic?Use the Arabic option in the app or device you type on. Mac and iPhone Dictation offer Arabic for Saudi Arabia and the United Arab Emirates, and Google Docs voice typing lists Arabic and 14 regional entries in Chrome, Edge or Safari. Windows 11 voice typing does not offer Arabic, so on Windows use Google Docs.
- How do you voice type in Bangla and get English text?The reliable method is two steps: voice type in Bangla, then translate the corrected Bangla text into English and read the result. One step is possible with OpenAI's open Whisper model, which translates Bangla speech into English, but its Bangla results were weak on the benchmark we checked, so use it only for a rough draft.
- How do you voice type in Chinese?On a Mac, add a Mandarin Chinese row in Dictation settings, switch to it before you speak, and check the output. On Windows 11, choose a Chinese voice typing language, then press Windows key and H in a text box, but expect voice typing there to need an internet connection. Choose the variant that matches the characters you write, and test one paragraph before you rely on it.
- How do you voice type in Hindi and get English text?To get English text from spoken Hindi, you need two steps: recognize the Hindi, then translate it. Each step adds its own errors, and a 2026 study of Vietnamese speech found that translating recognizer output instead of clean text lost quality. Check the Hindi transcript first, then the English.
- How do you voice type in Korean?On a Mac, Korean (Korea) is on Apple's on-device Dictation list, so add it in Dictation settings and switch to it before you speak. On Windows 11, Korean is a voice typing language, but voice typing needs an internet connection. The slips to watch for are word spacing and borrowed words, so read the output before you send it.
- How do you voice type in Russian?You can voice type in Russian with Apple's Dictation on a Mac, which lists Russian (Russia), and with Windows 11 voice typing, which lists one Russian language. On the Mac, Russian is also on the on-device list, so it can run without an internet connection. Windows voice typing needs an internet connection.
- How do you voice type in Urdu and get English text?Voice type in Urdu first, then translate the corrected Urdu text into English and read the result. Urdu is on Google's keyboard and Docs voice typing lists, but not on Windows voice typing. Whisper can translate Urdu speech into English in one step, but its results on a read-speech benchmark were moderate, so check every translation.
- How do you voice type in Urdu?Urdu voice typing has fewer built-in options than Hindi. Google announced Urdu voice input in 2017, Microsoft's Azure Speech table lists Urdu (India), and Windows 11 voice typing does not list Urdu. Urdu's error rate on a standard benchmark is higher than English, so check each result.
- How fast do people type in Japanese?The best measured figures for Japanese typing are about 17 words per minute on a phone and 23 on a desktop PC, from a small study of copying by university students. Japanese speed is usually counted in characters, and the 2nd level of a Japanese certification test requires about 60 characters per minute. We found no recent measured figure for typical Japanese typists, and no measured figure for Korean.
- How good is Whisper at Hindi?Whisper is usable in Hindi but noticeably weaker than in English. On the FLEURS benchmark the large-v2 model had a word error rate of 21.5 percent in Hindi, against 4.2 percent in English, and a fine-tuned Hindi model did better on a closely related test.
- How good is Whisper at Japanese?Whisper is accurate on clean, read Japanese with its large models: the large-v2 model had a character error rate of 5.3 percent on the FLEURS benchmark, and the smallest model did far worse. Set the language to Japanese yourself, use the largest model your app offers, and test with your own recordings before you rely on it.
- How good is Whisper at Korean?Whisper is usable in Korean but clearly weaker than in English. The large-v2 model had a word error rate of 14.3 percent on the FLEURS benchmark, about 3.4 times its English rate, and Korean's loose spacing makes word counts less exact. Set the language to Korean and test with your own speech before you rely on it.
- How many languages does Google Speech-to-Text support?Google's Cloud Speech-to-Text product page says the service supports over 125 languages. Its supported-languages table lists codes by region and by model, so one language often appears under several codes, and each model supports a different set. Find your exact language, locale and model in the table before you build on it.
- Is dictation good for language learning?Yes, for listening. The strongest controlled study we found, a 2002 trial in Tehran, reported a significant listening gain after frequent dictation. We found no controlled study of dictation for spelling or grammar, and the evidence for speaking comes from other kinds of practice.
- Is speaking Chinese faster than typing pinyin?In one controlled phone study, yes, for short messages. Mandarin entered by speech reached 123 words per minute, against 43 on a built-in pinyin keyboard, which is 2.87 times faster. The study used short messages in a lab, so test your own writing before you decide.
- Is speech recognition less accurate in languages other than English?Often, yes. Accuracy varies widely from one language to the next, and English is not the reference point. In one benchmark, Whisper's error rate in Yoruba was about 31.6 times its rate in Spanish, while English was only about 1.4 times Spanish.
- Is speech to text as accurate in Spanish as in English?On one published benchmark, Spanish scored better than English: Whisper large-v2 had a 3.0 percent word error rate in Spanish and 4.2 percent in English, on read speech. That benchmark does not measure dictation apps, and accuracy varies by dialect and by how you speak, so test the variant you use before you rely on it.
- Should you set dictation to US or UK English?Set the English variant closest to how you speak, not the spelling you prefer. No study we found compares US and UK settings in Apple's or Microsoft's dictation, so test both on the same passage and keep the one with fewer errors. Check the commands you use, because some are available only in U.S. English.
- What are difficult words for an English dictation test?The hardest words for an English dictation test are those where spelling does not match sound: silent letters as in knife, double letters as in necessary, vowel teams as in receive, and homophones such as principal and principle. Mix a few of them into each test rather than making every word hard.
- What are good dictation words for nursery, LKG and UKG in English?For nursery, use spoken dictation with short words such as cat, sun and cup, matched to pictures. For LKG, use short words whose spelling follows their sounds, such as bed, pen and mop. For UKG, add blends such as frog and stop, long vowels such as cake, and a few irregular words such as the and said.
- What are good Hindi dictation words for classes 1 to 4?In class 1, use short words made only of consonants with their inherent vowel, such as घर and कमल. In classes 2 to 4, add one matra at a time, then conjuncts such as स्कूल and the danda that ends a sentence. Check every list against the textbook the child is using.
- What are good Hindi dictation words without matra?Good words are those where every letter is a consonant with its inherent vowel, or an independent vowel letter, with no vowel sign attached. Examples include कमल, जल, घर, सच and नमक. Leave halant words for later, and check the spoken form, because a final vowel is sometimes not pronounced.
- What is a good Hindi typing speed?A good Hindi typing speed is about 30 words per minute, counted as correct words only. That is the Hindi criterion in CBSE's typing test, and two Jharkhand High Court notices asked for 35 and 25. No survey we could find measures what typical Hindi typists achieve, so use 30 as a target and time yourself on text you would really write.
- What is dictation called in Hindi?In Hindi, dictation is called श्रुतलेखन (śrutlekhan). The HinKhoj dictionary gives the shorter spelling श्रुतलेख, so you may see either one.
- What is the best Arabic speech to text API?There is no single best Arabic speech to text API, because the providers list different Arabic varieties and no shared test ranks them on dialect speech. Choose the variety you speak, check each provider's data terms, and test two or three of them on your own recordings.
- What is the best dictation app for learning a language?For listening and spelling, use a listen-and-type exercise that plays audio, lets you replay it and checks what you type. For pronunciation, use built-in speech to text in the language you are learning. No single app is best for every language, so first check that the tool supports your language and region.
- What is the best free Hindi speech to text?There is no single best free option. For dictating as you type, the built-in tool on your device is the best choice: Gboard on Android, Apple Dictation on iPhone, Windows voice typing, or Google Docs voice typing. For a recording, use a free transcription service with a stated allowance, and check the Hindi output, because Hindi error rates are higher than English.
- What is the best speech to text for Japanese?For most people, the simplest choice for Japanese speech to text is the built-in tool already on their device. Apple's Dictation on a Mac lists Japanese for on-device processing, and Windows voice typing supports Japanese but needs an internet connection. For a free model that runs on your own computer, Whisper is the best-known option, but we found no published test that compares it head to head with the built-in tools on the same Japanese audio.
- What languages does Dragon NaturallySpeaking support?The newest Dragon language list we could verify is for Dragon Professional Individual version 15. It covers English in several variants, plus Dutch, French, German, Italian and Spanish. We found no list for Dragon Professional 16 or for a Japanese edition, so confirm the languages for your exact edition with Nuance before you buy.
- What words should you use for an English dictation test for kids?Use words the children already know from class, then add a few less familiar words that they can spell by sound. Keep each list short and tied to the week's lesson. Score each spelling error once, and do not take marks off for a word that is misspelled again and again.
- Which APIs combine speech to text and translation?Azure Speech's speech translation service handles speech to text translation of an audio stream in one service. Google Cloud needs two calls: Speech-to-Text returns a transcript, and Cloud Translation translates that text. OpenAI's whisper-1 translation endpoint returns English from a complete recording, and Gemini can translate speech it is given as audio.
- Which app translates speech to text live?For live speech translated into text on a phone, use a translator app built for conversation: Google Translate, Apple Translate or Microsoft Translator are the main options we checked, and Samsung Interpreter works on Galaxy phones for the languages it lists. Microsoft Translator's auto mode allows 20 seconds per speaker on Android, and Teams live translated transcription needs an organizer with a Premium license. For a video or recording, transcribe it first and then translate the text.
- Which dictation app supports the most languages?Among the dictation apps and keyboards we could check, Google's Gboard on Android states the largest count we could verify, 119 language varieties, but that figure dates from 2017 and counts regional varieties. Apple publishes its list by language and region, and we found no current figure from a dictation app that you can compare directly, so check the list for your exact language and region.
- Which Spanish should you choose for dictation?Choose the variant that matches how you speak, and make sure your device lists it. Apple's Mac lists five Spanish variants for on-device dictation, and Windows 11 voice typing lists Spanish (Mexico) and Spanish (Spain). We found no published dictation test that ranks the variants, so test the one you pick with your own voice.
- Which speech to text API works best for Indian languages?No API has been shown to be best for every Indian language, so compare them on your own audio. Google Cloud lists a Hindi model, Azure lists Bengali and Urdu locales for India, and AI4Bharat publishes an open MIT-licensed Hindi model. Measure word error rate on your own speech before you choose.
- Which speech to text works best with an Indian accent?In the 2023 Svarah test of Indian English, the large open source Whisper model had the lowest word error rate of the systems tested, at 7.2 percent. The test is from 2023 and covers a limited set of speakers, and results differ by first language, so run the same passage through the options you can use before you choose.
- Why does dictated Arabic or Hebrew text look backwards?Dictated Arabic or Hebrew is usually not stored backwards. Unicode stores it in logical order, and the screen reorders each run of right-to-left and left-to-right text, so punctuation, parentheses and numbers at the boundaries can end up on the side you did not expect. Proofread those boundaries first.
- Why does it say dictation is not available in the current language?The message means voice typing has no support for the language your PC is using for input, so Windows will not listen. Switch to a supported language with Windows logo key + Spacebar, or add and install a language under Time & language. If your language is not on Windows' list, no setting will create support for it, so use one of the other options below.
- Why does Whisper transcribe in the wrong language?Whisper transcribes in the wrong language for one of four main reasons: it guessed from the first 30 seconds, the speech mixes languages, the language had little training data, or the translate task produced English. Setting the language yourself is the first fix to try.
- Why doesn't voice recognition understand my accent?Most dictation systems are trained mainly on speech that is typical for their training data, so an accent they heard little of produces more errors. The gap is real and measurable, but accent is not the only cause. A distant microphone, fast speech and background noise cause errors too, and those are often easier to fix.
- Why is Google Translate voice not working?Google's help points to four causes when Google Translate voice fails: the spoken language cannot be translated from speech, so the Speak or Dictation button is disabled; the microphone permission is off; the browser does not support microphone translation; or the spoken language is set wrong. Check the button first, then the permission, then the browser.
Terms in this topic
- AccentAn accent is the way a speaker pronounces a language, shaped by where they grew up, their first language or their community. Accents change pronunciation, not vocabulary, and speech recognition error rates can differ sharply from one accent to another.
- Character error rate (CER)Character error rate (CER) is the number of character insertions, substitutions and deletions needed to turn a recognizer's output into the correct text, divided by the number of characters in the correct text. It is the usual measure where words are not separated by spaces.
- Code-switchingCode-switching is the habit of moving between two or more languages within a conversation or a single sentence. Speech recognizers built around one language often misread the switched words, so mixed speech is harder to dictate than speech in one language.
- DiacriticsDiacritics are marks added to a letter, such as the acute accent in é, the tilde in ñ or the vowel marks in Arabic, that change how a letter sounds or which word it spells. Speech tools and everyday writers often leave them out, so a dictated word can come out spelled differently from what you meant.
- DialectA dialect is a regional or social variety of a language, with its own vocabulary, grammar and pronunciation. Speech recognition is usually less accurate on dialects than on the standard form, because labeled training data for dialects is often scarcer.
- Input method editor (IME)An input method editor (IME) is software that supports text input for characters that cannot be typed easily on a standard keyboard, such as many East Asian characters. Speech input can also be an input method, and Android allows only one input method to be enabled at a time.
- LocaleA locale is a language combined with a region, and sometimes a script, such as en-GB for English as used in the United Kingdom. Dictation and speech settings use locales because the same language can differ in spelling, punctuation and speech patterns from one region to another.
- Low-resource languageA low-resource language is one with little digital speech and text data for training speech recognition and other language tools. There is no fixed cutoff, and a language can be well resourced for one task and poorly resourced for another.
- Modern Standard Arabic (MSA)Modern Standard Arabic (MSA), also called Fusha, is the standardized form of Arabic used in writing and formal speech, such as news broadcasts and official documents. Speakers use regional dialects in everyday conversation, and speech tools recognize formal Arabic more accurately than dialect speech.
- Non-native speechNon-native speech is speech in a language that is not the speaker's first language, often called L2 speech or foreign-accented speech. Recognizers can transcribe read non-native English well, but they handle hesitations, repetitions and revisions differently, and results depend on the system.
- PinyinPinyin, also called Hanyu Pinyin, is the system that spells Mandarin Chinese syllables with the Latin alphabet and marks the tones with diacritics. People type Chinese by spelling the sound in Pinyin and choosing the character they want, and speech tools must choose among characters that share one Pinyin spelling.
- Right-to-left (RTL) scriptA right-to-left (RTL) script is a writing system, such as Arabic, Hebrew, Persian or Urdu, whose letters run from right to left. Text that mixes these letters with digits or Latin words is bidirectional, so software needs a specific algorithm to order and display it.
- Speech translationSpeech translation converts speech in one language into the same meaning in another language, as text or as synthesized speech. Most systems first recognize the speech and then translate the text, so a recognition error carries into the translation.
- Spoken language identificationSpoken language identification is the task of working out which language is being spoken in a recording. Products call it automatic language detection. It usually chooses from a short list of candidate languages and decides for a segment of audio rather than word by word.
- TransliterationTransliteration represents the text of one language in the writing system of another, for example by writing a word in Latin letters. It keeps the spelling or sound and does not translate the meaning, which is what separates it from translation.
Referenced by
- Dictation and Voice Typing: How to Talk Instead of TypeTopic
- Mac Dictation: How to Dictate and Control a Mac by VoiceTopic
- Dictation on Windows: Voice Typing, Voice Access, AppsTopic
- Dictation and Voice to Text on iPhone and AndroidTopic
- Speech Recognition: What It Is and How Well It WorksTopic
- Transcription: How to Turn Audio and Video Into TextTopic
- Dictation Languages: What Each App and Device SupportsGuide
- How to Change the Dictation Language on Any DeviceGuide
- Hindi Voice Typing and Speech to TextGuide
- Speech to Text With Translation on Phones and ComputersGuide
- Speech Recognition and Accents: Accuracy and FixesGuide
- Multilingual Dictation: Two Languages, Auto-Detect, MixingGuide
- Spanish Speech to Text and Talk to TextGuide
- Japanese, Chinese and Korean Speech to TextGuide
- How to Translate Audio and Video to English TextGuide
- Dictation for Language Learning: How and Why It WorksGuide
- Dictation Languages on iPhone and AndroidGuide
- Arabic Speech to Text and Right-to-Left DictationGuide
- Speech to Text in French, German, Russian and MoreGuide
- Can Whisper translate audio?Question
- Can you dictate in two languages at the same time?Question
- Can you train dictation software to your accent?Question
- Can you translate Zulu or Xhosa audio to English?Question
- Can you use voice typing to practice pronunciation?Question