Library
Facts and numbers
4,987 facts. Each one has its own page with the exact words of its source, what it measures and what it does not show.
Dictation
390 facts. The most cited:
- 52 words per minuteIn the same controlled study on an iPhone, English text entered on the built-in QWERTY keyboard reached 52 words per minute, against 153 words per minute by speech.Sherry Ruan et al., Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT), 2018
- Windows voice typing needs an internet connection, a working microphone, and the cursor in a text box.Microsoft, 2026
- Apple says Dictation on iPhone is processed on the device in many languages, and that no internet connection is required.Apple Inc., Apple Support, iPhone User Guide, n.d.
- 3 browsersGoogle says voice typing in Google Docs works with the latest versions of Chrome, Edge and Safari, and that the web browser controls the speech-to-text service.Google (Docs Editors Help), 2026
- On an iPhone, you turn on Dictation in the Settings app under General, then Keyboard, by turning on Enable Dictation and confirming the prompt.Apple Inc., Apple Support, iPhone User Guide, n.d.
- On Android with Gboard, you voice type by tapping a text field, tapping the microphone at the top of the keyboard and speaking when you see "Speak now".Google (Gboard Help), 2026
- 30 secondsMac Dictation stops automatically when it detects no speech for 30 seconds.Apple Inc., Apple Support, Mac User Guide, n.d.
- In Google Docs, you start voice typing by opening a document in a supported browser, choosing Tools then Voice typing, and clicking the microphone box that appears.Google (Docs Editors Help), 2026
- Windows 11 Voice Access controls the PC and writes text by voice without an internet connection.Microsoft Support, 2026
- Apple says dictation sent to its servers is not stored unless you opt in to Improve Siri and Dictation.Apple, Apple, 2026
Dictation software
342 facts. The most cited:
- Apple says that, on a Mac, voice inputs and transcripts for general text Dictation are processed on your device and not sent to Siri servers, and that you can check this in Keyboard settings.Apple Inc., Apple Support, Mac User Guide, n.d.
- 10.3 percent word error rateIn a 2025 9to5Mac test of one 7-minute-31-second episode, Apple's speech API had a word error rate of 10.3 percent, against 1.5 percent for Whisper Large V3 Turbo.Marcus Mendes, 9to5Mac, 2025
- 1.5 percent word error rateIn the same 2025 9to5Mac test, Whisper Large V3 Turbo had a word error rate of 1.5 percent on the episode, the lowest of the three engines the author tested.Marcus Mendes, 9to5Mac, 2025
- 22 October 2018 date updates endedNuance said it would stop providing updates for Dragon Professional Individual for Mac after 22 October 2018.Dictate.com.au, Dictate.com.au (news and blog, reseller), 2018
- 685.99 US dollarsCDW lists a one-user Dragon Professional (v. 16) license at $685.99 on the day we checked.CDW, n.d.
- On Superwhisper's Free plan, dictation uses local Whisper models without AI processing, and the company says nothing dictated leaves the device.Superwhisper, Superwhisper documentation, 2026
- In the HyPoradise benchmark, a large language model can correct words that are missing from the recognizer's N-best list, because it generates text rather than only choosing among candidates.Chen Chen et al., arXiv (NeurIPS 2023 dataset track), 2023
- The Handy project describes itself as a free, open source speech-to-text application that works completely offline.CJ Pais, GitHub, 2026
- 2,000 words per week on desktopWispr Flow's free plan allows 2,000 dictated words per week on desktop and 1,000 per week on mobile, according to its pricing page.Wispr Flow, Wispr Flow (vendor website), 2026
- Buzz describes itself as transcribing and translating audio offline on a personal computer.Chidi Williams, GitHub (Buzz repository), 2026
Typing and writing speed
518 facts. The most cited:
- 153 words per minuteIn a controlled study on an iPhone, English text entered by speech reached 153 words per minute, against 52 words per minute on the built-in keyboard, 2.93 times faster.Sherry Ruan et al., Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT), 2018
- 1.30% uncorrected error rate with speechSpeech produced fewer errors during entry than the phone keyboard (5.30% vs 11.22% corrected error rate) but left slightly more errors in the final text (1.30% vs 0.79% uncorrected error rate).Sherry Ruan et al., Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT), 2018
- 51.56 words per minuteIn a study of 136 million keystrokes from an online typing test, the average participant typed 51.56 words per minute, with a standard deviation of 20.2.Vivek Dhakal et al., CHI 2018, ACM Conference on Human Factors in Computing Systems, 2018
- 123 words per minuteIn the same study, Mandarin Chinese entered by speech reached 123 words per minute, against 43 words per minute with the Pinyin keyboard, 2.87 times faster.Sherry Ruan et al., Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT), 2018
- 36.2 words per minuteIn an online study of 37,370 volunteers typing on their own phones, the average speed was 36.2 words per minute, with 2.3 percent uncorrected errors.Kseniia Palin et al., Proceedings of the 21st International Conference on Human-Computer Interaction with Mobile Devices and Services (MobileHCI 2019), 2019
- 196 words per minuteIn large corpora of conversational telephone speech in English and Chinese, the overall average rate was 196 words per minute, counting all elapsed time.Jiahong Yuan et al., Interspeech 2006, 2006
- 78 words per minuteIn a study of 136 million keystrokes from an online typing test, typists faster than 90 percent of participants were those above about 78 words per minute.Vivek Dhakal et al., CHI 2018, ACM Conference on Human Factors in Computing Systems, 2018
- The motion capture study found three predictors of high performance: a letter is always pressed by the same finger, upcoming keystrokes are actively prepared, and hand movement is kept small.Anna Maria Feit et al., Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 2016
- The authors write that touch typing can be learned quickly but needs deliberate training of hundreds of hours to reach the performance rates reported in the literature.Anna Maria Feit et al., Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 2016
- 26 words per minuteIn the same online study, the slowest 10 percent of typists were those below about 26 words per minute.Vivek Dhakal et al., CHI 2018, ACM Conference on Human Factors in Computing Systems, 2018
Dictation on Mac
247 facts. The most cited:
- Apple says that with Dictation on a Mac you can speak to enter text anywhere you can type it.Apple Support, n.d.
- 63 language variantsApple's macOS feature availability page lists 63 language and regional variants for Dictation on the Mac.Apple, Apple, 2026
- On a Mac you start Dictation by pressing the Microphone key (on keyboards that have one), pressing the Dictation keyboard shortcut, or choosing Edit > Start Dictation.Apple Inc., Apple Support, Mac User Guide, n.d.
- Current Mac Dictation accepts text of any length without a timeout; you stop it with Escape, the Microphone key or the Dictation shortcut.Apple Inc., Apple Support, Mac User Guide, n.d.
- On a Mac, when Voice Control is on, you dictate through Voice Control and standard macOS Dictation is not available.Apple Support, n.d.
- With more than one dictation language set up, you switch languages by clicking the language label next to the cursor or by pressing the Globe key and choosing a language.Apple Inc., Apple Support, Mac User Guide, n.d.
- 48 language variantsApple lists 48 language and regional variants for on-device and modeless Dictation on the Mac, which require a download of speech models.Apple, Apple, 2026
- Apple says you can change the language you dictate in, or set up Dictation for multiple languages and switch between them as needed. Apple describes switching between languages, not recognizing two languages within one sentence.Apple Inc., Apple Support, Mac User Guide, n.d.
- In supported languages, Mac Dictation inserts commas, periods and question marks automatically, and you can switch this off with the Auto-punctuation setting in Keyboard settings.Apple Inc., Apple Support, Mac User Guide, n.d.
- For people who opt in to Improve Siri and Dictation, Apple says only its own employees, under confidentiality obligations, can access the audio.Apple, Apple, 2026
Dictation on Windows
315 facts. The most cited:
- On Windows 11, you turn on voice typing by pressing the Windows key + H on a hardware keyboard, or the Microphone button on the touch keyboard.Microsoft, 2026
- Microsoft says that if Voice Access is running and you press the Windows key + H, voice typing does not start and Windows tells you to dictate with Voice Access instead.Microsoft, Microsoft Support, 2026
- Microsoft says Voice Access setup downloads language files for the chosen language to be used by on-device speech recognition, and that this step needs an internet connection.Microsoft Support, n.d.
- Microsoft says Voice Access lets you control your PC and write text by voice without an internet connection.Microsoft, Microsoft Support, 2026
- 46 languages and variantsMicrosoft lists 46 languages and regional variants that support voice typing in Windows 11, from Bulgarian to Vietnamese, including six variants of English.Microsoft Support, 2026
- Microsoft describes fluid dictation as a new Voice Access feature on Copilot+ PCs, so it is not available on every Windows 11 PC.Microsoft Support, 2026
- Microsoft's troubleshooting steps for voice typing start by asking you to select the microphone you want to use in Settings, under System, Sound, Input.Microsoft, Microsoft Support, n.d.
- 15 languages and dialectsMicrosoft lists 15 languages and dialects for Voice Access in Windows 11: six English variants, two Spanish, German, two French, Simplified and Traditional Chinese, Japanese and Italian.Microsoft Support, n.d.
- Microsoft says Voice access offers both commanding and dictation, so it can control every task on the PC, not only write text.Microsoft, Microsoft Support, 2026
- Microsoft says Voice Access uses on-device speech recognition and works even without the internet.Microsoft, Microsoft Support, 2026
Accessibility and health
508 facts. The most cited:
- The U.S. National Library of Medicine says studies have not proved that typing on a computer, using a mouse or other repeated movements cause carpal tunnel syndrome.MedlinePlus, U.S. National Library of Medicine, 2025
- 6 weeksThe NHS advises wearing a wrist splint at night to relieve pressure on the nerve in carpal tunnel syndrome, and says it may take up to 6 weeks before symptoms start to improve.NHS (National Health Service, England), 2024
- Carpal tunnel numbness and pain are mainly felt in the thumb, index, middle and ring fingers, and often wake people at night.American Academy of Orthopaedic Surgeons, OrthoInfo (AAOS), n.d.
- Windows Voice Access has an Add to vocabulary command that Microsoft lists for improving recognition of complex or hard to pronounce words.Microsoft Support, 2026
- If carpal tunnel syndrome is left untreated for too long, it can lead to permanent dysfunction of the hand, including loss of finger sensation and weakness.American Academy of Orthopaedic Surgeons, OrthoInfo (AAOS), n.d.
- Carpal tunnel syndrome is excessive pressure on the median nerve at the wrist, the nerve that gives feeling and movement to parts of the hand.MedlinePlus, U.S. National Library of Medicine, 2025
- The NHS says that if carpal tunnel syndrome is getting worse and other treatments have not worked, surgery may be discussed, and that surgery usually cures it.NHS (National Health Service, England), 2024
- Google states that automatic speech recognition systems are most often trained on typical speech, so people with speech impairments or heavy accents do not get the same benefit.Google Research, Google Research blog, 2019
- In a 1995 study of 29 university students with learning disabilities, essays written with a speech recognition system received significantly higher holistic scores than essays written without assistance.Eleanor L. Higgins and Marshall H. Raskind, Learning Disability Quarterly (ERIC record EJ516141), 1995
- In a laboratory study, the activity of typing raised pressure in the carpal tunnel compared with holding the same hand posture still.David M. Rempel et al., Journal of Orthopaedic Research, 2008
Dictation for work and study
291 facts. The most cited:
- Windows voice typing needs an internet connection, a working microphone and the cursor in a text box before it will type what you say.Microsoft, 2026
- On Windows, you start voice typing by pressing the Windows key and H together on a hardware keyboard.Microsoft, 2026
- The US Bureau of Labor Statistics says medical transcriptionists listen to the recorded dictation of physicians and other healthcare workers and transcribe it into medical reports.U.S. Bureau of Labor Statistics, 2025
- 12.7 minutes saved per 1,000 wordsAt the study's 153 words per minute by speech and 52 on the keyboard, a 1,000-word draft takes about 6.5 minutes spoken and 19.2 minutes typed, a saving of about 12.7 minutes.Calculated from other facts
- BLS lists reviewing and editing drafts prepared by speech recognition software among the usual duties of medical transcriptionists.U.S. Bureau of Labor Statistics, 2025
- The BLS says advances in speech recognition and natural language processing let physicians document patient encounters in real time, reducing the need for medical transcriptionists.U.S. Bureau of Labor Statistics, 2025
- 1.47 percent hallucination rateIn the npj Digital Medicine framework study, LLM-generated clinical notes had a hallucination rate of 1.47 percent, measured against clinician-annotated sentences.Asgari E et al., npj Digital Medicine, 2025
- 11.5 percent of reviewed notes with a hallucinationIn the UC Davis pilot, hallucinations, meaning content the recording did not support, appeared in 11.5 percent of the 356 reviewed notes, the second most common error after omissions.Taylor SL et al., JMIR Medical Informatics, 2026
- 18 percent of reviewed notes with an accidental omissionIn the UC Davis pilot, accidental omissions were the most frequent error, in 18 percent of the 356 reviewed notes, ahead of hallucinations at 11.5 percent and accidental inclusions at 9.3 percent.Taylor SL et al., JMIR Medical Informatics, 2026
- Under HIPAA, a business associate is a person who, other than as a member of the covered entity's workforce, creates, receives, maintains or transmits protected health information.Legal Information Institute, Cornell Law School (text of the Code of Federal Regulations), n.d.
Dictation in your apps
372 facts. The most cited:
- OpenAI says that when you press the microphone icon in ChatGPT, the recorded audio is sent to its models to be transcribed, and the returned text can be edited before you send it.OpenAI, OpenAI Help Center, 2026
- Microsoft says the Dictate feature in the Word desktop app is only available to Microsoft 365 subscribers.Microsoft, 2026
- Microsoft says the dictation feature in Outlook is only available to Microsoft 365 subscribers.Microsoft, Microsoft Support, 2026
- 10 GB of video memoryThe large Whisper model needs about 10 GB of video memory according to OpenAI's model table.OpenAI, GitHub (OpenAI repository), 2023
- By default, speech recognition on a web page sends your audio to a server-based recognition service, so it does not work offline.MDN contributors, MDN Web Docs (Mozilla), 2026
- Claude Code's voice dictation streams the recorded audio to Anthropic's servers for transcription, and the audio is not processed on the computer.Anthropic, Anthropic, 2026
- OpenAI's speech-to-text guide names whisper-1 as the model to use for translating a completed audio recording into English.OpenAI, OpenAI, 2026
- Microsoft says Dictation in Office needs a microphone and a reliable internet connection.Microsoft Support, 2026
- Chrome's 139 release notes say the Web Speech API gained on-device speech recognition, so a website can make sure neither audio nor transcribed speech is sent to a third-party service.Google Chrome for Developers, 2025
- OpenAI's ChatGPT Voice help article says to use ChatGPT Dictation when you want to record a prompt, review and edit its transcription, and then send it as text.OpenAI, OpenAI Help Center, 2026
Dictation on phones and tablets
284 facts. The most cited:
- Samsung says Galaxy phones and tablets have two built-in voice-to-text apps, Samsung voice input and Google Voice Typing, and Samsung voice input works only with the Samsung Keyboard.Samsung US Support, 2026
- Google's App Store listing for Gboard on iPhone says that voice input from the Gboard microphone is sent to Google for transcription.Apple App Store, n.d.
- Apple says that when you use a keyboard for another language, you can use Dictation in that language as well.Apple Inc., Apple Support, iPhone User Guide, n.d.
- Apple says Dictation on iPhone may not be available in all languages or in all countries or regions, and that features may vary.Apple Inc., Apple Support, iPhone User Guide, n.d.
- In February 2025, 9to5Google reported that Gboard for Android features offline voice typing on all devices.9to5Google, 2025
- With Apple Intelligence in iOS 27, iPhone dictation uses an on-device model to improve spelling, punctuation and capitalization, in English on iPhone 17 Pro, iPhone 17 Pro Max, iPhone Air or later.Apple Inc., 2026
- In Gboard's voice toolbar, you switch languages by tapping the abbreviation at the bottom left, which shows your default language.Google (Gboard Help), 2026
- Apple says that with Dictation on iPhone you can dictate text in any place where you can type it.Apple Inc., Apple Support, iPhone User Guide, n.d.
- Samsung's keyboard settings have an Offline voice input section where you install offline language packs.Samsung US Support, 2026
- Apple's warning on third-party keyboards says that if you enable Full Access, developers are permitted to access, collect and transmit the data you type.Matthew Panzarino, TechCrunch, 2014
Speech recognition
436 facts. The most cited:
- 80 MBIn 2019 Google moved Gboard voice typing onto the phone with an 80 MB speech recognizer that works even when the phone is offline, first on Pixel phones in American English.Johan Schalkwyk, Google Research Blog, 2019
- 7.44 percent average WEROn the same leaderboard track, OpenAI Whisper Large v3 had an average word error rate of 7.44 percent and an inverse real-time factor of 146.Hugging Face and collaborators, arXiv, 2025
- 6.05 percent average WEROn the Open ASR Leaderboard's English short-form track, NVIDIA Parakeet TDT 0.6B v2 had an average word error rate of 6.05 percent and an inverse real-time factor of 3,390.Hugging Face and collaborators, arXiv, 2025
- 2 GBGoogle said its traditional server-side recognizer used a search graph of almost 2 GB, too large for a phone, which is why it needed a connection.Johan Schalkwyk, Google Research Blog, 2019
- A 2025 study of Whisper found that when non-speech sounds were played to the model, a set of hallucinated phrases appeared frequently.Mateusz Barański et al., arXiv (accepted for IEEE ICASSP 2025), 2025
- OpenAI's Whisper package requires the command-line tool ffmpeg to be installed on your system, in addition to the Python package.OpenAI, GitHub (OpenAI repository), 2023
- 98 languagesOpenAI's speech to text guide says Whisper supports 98 languages, and that accuracy varies by language.OpenAI, OpenAI, 2026
- Google says Cloud Speech-to-Text does not log customer audio or transcripts unless the customer opts into data logging.Google Cloud, 2026
- The Careless Whisper study evaluated Whisper as it stood in 2023, so its 1 percent hallucination figure describes that version of the model.Allison Koenecke et al., arXiv (ACM FAccT 2024), 2024
- 30 secondsThe Whisper paper says its models are trained on 30-second audio chunks and cannot take longer audio as a single input.Alec Radford et al., OpenAI (arXiv), 2022
Transcription
336 facts. The most cited:
- 1 percentIn a 2024 study, roughly 1 percent of Whisper audio transcriptions contained entire hallucinated phrases or sentences that did not exist in the audio.Allison Koenecke et al., arXiv (ACM FAccT 2024), 2024
- 25 MBOpenAI's Transcriptions API accepts audio files up to 25 MB, and larger recordings must be compressed or split into chunks.OpenAI, OpenAI, 2026
- A 2025 study on dementia speech says that Whisper tends to omit filler words such as uh and um from its transcripts.Emmanuel Akinrintoyo et al., arXiv, 2025
- $0.0045 per minuteOpenAI's price list shows its gpt-transcribe model at $0.0045 per minute and gpt-4o-transcribe at $0.006 per minute.OpenAI, OpenAI, 2026
- The Federal Trade Commission says that if someone asks you to pay to get a job, that is a scam, and honest employers will never ask you to.Federal Trade Commission, FTC Consumer Advice, 2026
- The authors say that spontaneous switching within a sentence causes language confusion for code-switching recognizers, and that accent bias blurs the sound boundaries between languages.Hexin Liu et al., arXiv, 2025
- A federal class action filed in August 2025 alleges that Otter.ai's Otter Notebook service, by default, does not ask meeting attendees for permission to record.NPR, 2025
- $1.99 per minuteRev's price page lists human transcription with a 99 percent plus accuracy guarantee in 12 hours or less at $1.99 per minute.Rev, Rev, 2026
- 442 timesAt $0.0045 per minute an hour of audio costs about $0.27 through OpenAI's API, against about $119.40 at Rev's $1.99 per minute for human transcription, roughly 440 times more.Calculated from other facts
- 38 percentThe study's thematic analysis found that 38 percent of Whisper's hallucinations included explicit harms such as perpetuating violence, making up inaccurate associations or implying false authority.Allison Koenecke et al., arXiv (ACM FAccT 2024), 2024
Voice privacy and security
286 facts. The most cited:
- California Penal Code 632 punishes a person who intentionally records a confidential communication without the consent of all parties.California Public Law, 2026
- HHS says the covered entity and a cloud service provider that holds its health information must sign a HIPAA-compliant business associate agreement.U.S. Department of Health and Human Services, Office for Civil Rights, n.d.
- HIPAA's Privacy Rule lets a covered entity allow a business associate to create, receive, maintain or transmit protected health information on its behalf only if the entity obtains satisfactory assurance that it will be safeguarded.Legal Information Institute, Cornell Law School, 2026
- The HIPAA Security Rule makes an accurate and thorough risk analysis of the risks to the confidentiality, integrity and availability of electronic protected health information a required implementation specification.Legal Information Institute, Cornell Law School, 2026
- The federal Wiretap Act says it is not unlawful for a person to intercept a wire, oral or electronic communication when they are a party to it or one party has given prior consent to the interception.Legal Information Institute, Cornell Law School, 2026
- ABA Formal Opinion 512 concludes that informed consent is required before client information is entered into self-learning generative AI tools whose outputs could disclose it.Harvard Journal of Law and Technology (JOLT Digest), 2026
- Illinois's Biometric Information Privacy Act defines biometric identifiers to include voiceprints, alongside fingerprints and iris scans.Hogan Lovells, Hogan Lovells, 2019
- Microsoft documents that for real-time speech to text in Azure, audio is processed only in server memory and no data is stored at rest.Microsoft, Microsoft Learn, 2026
- 30 daysOpenAI says abuse monitoring logs are generated for all API feature usage and retained for up to 30 days by default, unless longer retention is required by law or to protect its services.OpenAI, OpenAI Platform documentation, 2026
- Article 9 of the GDPR prohibits processing special categories of personal data, including biometric data used for the purpose of uniquely identifying a natural person, unless one of the listed exceptions applies.gdpr-info.eu (Intersoft Consulting), 2016
Languages and accents
337 facts. The most cited:
- 4.2 percent word error rateOn the Fleurs benchmark, the Whisper large-v2 model had a word error rate of 4.2 percent for English.Alec Radford et al., OpenAI (arXiv), 2022
- 21.5 percent word error rateOn 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
- 3.0 percent word error rateOn 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
- 39.3 percent word error rateOn the Fleurs benchmark, the Whisper large-v2 model had a word error rate of 39.3 percent for Swahili.Alec Radford et al., OpenAI (arXiv), 2022
- Microsoft's list of languages that support voice typing in Windows 11 includes Hindi, Japanese, Korean, Marathi and Italian (Italy).Microsoft, 2026
- 46 to 71 percent relative word error reductionIn a 2007 study of foreign-accented English, the best combination of pronunciation modelling and acoustic adaptation cut word error rate by 46 to 71 percent relative.Ghazi Bouselmi et al., arXiv, 2007
- 8.3 percent word error rateOn the Fleurs benchmark, the Whisper large-v2 model had a word error rate of 8.3 percent for French.Alec Radford et al., OpenAI (arXiv), 2022
- OpenAI says Whisper's performance varies widely depending on the language, and it publishes a per-language breakdown of error rates.OpenAI, GitHub (OpenAI repository), 2023
- A 2025 review of code-switching speech recognition found that a small subset of languages receive the most attention and that there is a large disparity in the languages covered by datasets.Maha Tufail Agro et al., arXiv (accepted at LREC 2026), 2025
- 30 secondsWhen no language is set, Whisper's transcribe function detects the spoken language from up to the first 30 seconds of audio.OpenAI, GitHub (openai/whisper), 2026
Voice and AI
322 facts. The most cited:
- 8 secondsHome Assistant's documentation says Whisper takes around 8 seconds to process a voice command on a Raspberry Pi 4 and under a second on an Intel NUC.Home Assistant project, Home Assistant, 2026
- Home Assistant says Assist handles commands first by default, and only sends questions or commands it does not understand to the AI you have set up.Home Assistant project, Home Assistant blog, 2025
- 1 secondHome Assistant says Speech-to-Phrase, a closed-ended model that transcribes only what it knows, transcribes in under one second even on a Home Assistant Green or Raspberry Pi 4.Home Assistant project, Home Assistant, 2026
- 51.56 words per minuteIn a study of 136 million keystrokes from an online typing test, the average typing speed of participants was 51.56 words per minute.Vivek Dhakal et al., CHI 2018, ACM Conference on Human Factors in Computing Systems, 2018
- Google says that once a home switches to Gemini for Home voice assistant, it cannot switch back to Google Assistant.Google, n.d.
- Microsoft says fluid dictation in Windows Voice Access automatically corrects grammar, punctuation and filler words as you speak.Microsoft Support, 2026
- Microsoft says the fluid dictation experience in Windows Voice Access is powered by on-device small language models (SLMs).Microsoft Support, 2026
- In the 2025 INTERSPEECH paper, existing language-model error correction for speech recognition relies mainly on the text, ignores phonetic cues, and that leads to over-correction, meaning the model changes words that were already right.Natsuo Yamashita et al., arXiv (accepted by INTERSPEECH 2025), 2025
- Even when an LLM is asked only for grammar edits, the same 2026 preprint finds, it still changes the text in a way that significantly alters its semantic meaning.Marwa Abdulhai et al., arXiv, 2026
- A 2026 preprint reports that LLMs not only alter the voice and tone of human writing but also consistently alter its intended meaning.Marwa Abdulhai et al., arXiv, 2026
Other topics
3 facts. The most cited:
- Microsoft's voice typing steps say that on a hardware keyboard you turn voice typing on by pressing Windows key + H, with the cursor in a text box.Microsoft, 2026
- Microsoft's Windows 10 instructions say that you select a text field and press the Windows logo key + H to open the dictation toolbar.Microsoft, 2026
- Apple says you can send a copy of a Pages document on iPhone in another format, such as PDF, Word, EPUB, Plain Text or RTF.Apple Support, 2026