Guides

Dictation for academic writing on a Mac: papers, theses and grants by voice

Dictation for academic writing on a Mac works best for the prose: the argument of a paper, a thesis chapter, a literature review or the narrative of a grant. Say the sentences, and leave citations to your reference manager and markup to your keyboard. YapToText, the free app I built, types into Word, Google Docs, a .tex file or any other app, transcribes on your Mac, and has no silence timeout by default.

I built it because my hands make typing difficult, and I dictate every day. If you are writing a class essay, the guide for students fits better, and for fiction or a book, the guide for writers. This page is for researchers, PhD students and anyone writing for a committee, a journal or a funder.

Download YapToText on the Mac App StoreDownload for free. No account, nothing locked, macOS 14 or later on Apple silicon. Source on GitHub.

On this page: 11 sections
  1. Steps
  2. Draft a paper, thesis chapter or grant section by voice
    1. Step 1: Click where the section goes
    2. Step 2: Tap Right Command and talk one paragraph
    3. Step 3: Tap it again
  3. Keep your own words with Raw Transcription
    1. Step 1: Turn Auto mode off
  4. Teach it your field's terms and the authors you cite
  5. Leave citations to your reference manager
  6. LaTeX: say the prose, type the markup
    1. Step 1: Type the structure
    2. Step 2: Pick the mode for an editor
  7. Long writing sessions with no silence cutoff
    1. Keep every take until the thesis is done
  8. Transcribe interviews for qualitative research
  9. What it cannot do
  10. The short version
  11. Questions

Steps

  1. Turn off Auto mode on YapToText's Home page so Raw Transcription types your exact words.
  2. Add your field's terms and the authors you cite on the Dictionaries page.
  3. Click into your draft, tap Right Command, say one paragraph, and tap Right Command again.
  4. Say insert cite marker where a citation goes, then insert the citations with your reference manager.
  5. In Settings, Advanced, History & audio, set Keep to Keep everything and Auto-delete old items to Never.

Draft a paper, thesis chapter or grant section by voice

Academic prose is dense, but it is still sentences, and most people explain their research out loud more clearly than they type it. Say the section the way you would explain it to your advisor, then do the precise writing in the edit.

Step 1: Click where the section goes

  • You: Open the draft in Word, Google Docs, Scrivener or your editor, and click where the next paragraph goes.
  • YapToText: Nothing yet. It types into the app in front, never into its own window.
  • Check: The cursor is blinking in the draft.

Step 2: Tap Right Command and talk one paragraph

  • You: Tap Right Command once and say one paragraph, one claim or one step of the method. Press 2 while you talk if you want Raw Transcription for this dictation, in the default order.
  • YapToText: Opens the floating panel with your voice as a wave and your words as they form. A pause to think does not end it.
  • Check: The wave moves when you speak. If it stays flat, the app is not hearing your microphone yet.

Step 3: Tap it again

  • You: Tap Right Command again.
  • YapToText: Finishes the transcript on your Mac, pastes it at the cursor and puts back whatever you had copied before.
  • Check: The paragraph is in your draft. If it went to the clipboard instead, Accessibility is not on yet; on the Home page, press Grant next to Accessibility.

A few habits suit academic work in particular:

  • Outline out loud first. Talk through the section and press 4 while you talk, and Note mode gives you headings and bullets with nothing added. Then draft against the outline.
  • Leave a marker for every citation. Say the claim, drop a marker, and keep going. The section on citations below has a snippet for it.
  • Paste exact quotations. A quotation from a source has to match the source, so copy it in rather than saying it.
  • Say the breaks. Whisper punctuates ordinary speech by itself. Say new paragraph at the end of a sentence where a break belongs.

The guide to dictating well covers the habits in more depth, and if you write your thesis in Scrivener, dictation in Scrivener covers that app.

Keep your own words with Raw Transcription

A paper or a thesis has to be in your words, and many institutions, journals and funders now have rules about AI tools. I make no claims about any of them, and the rules can differ by department or by venue. Check the rules that apply to you, then pick the mode that fits them.

Here is exactly what touches a model in YapToText. Whisper, the speech model, turns your voice into words in every mode, the job any dictation tool's speech recognition does, and it runs on your Mac. Raw Transcription stops there: literal speech to text with basic punctuation, and no AI cleanup. Every other mode, Auto included, hands the transcript to an on-device cleanup model, Phi-3.5 Mini Instruct by default.

YapToText's Switch Mode list: Auto, Raw Transcription, Clean Up, Note, Email, Message and Code Comment, numbered 1 to 7.
Figure 1. Raw Transcription is the one mode tagged Verbatim. Every other mode, Auto included, is tagged AI cleanup.

Step 1: Turn Auto mode off

  • You: On the Home page, in the After transcription card, switch Auto mode off.
  • YapToText: Makes Raw Transcription the selected mode, until you pick another.
  • Check: After your next dictation, open it on the History page and click Show pipeline details. The Mode line reads Raw Transcription.

Two other routes get the same result. With Auto on, press 2 while you talk to use Raw Transcription for one dictation. And with Post-transcription analysis, the master switch on the Home page, turned off, every dictation is typed exactly as spoken. If your rules allow it later, History keeps each raw transcript, and Regenerate with AI can clean an old take up without dictating it again. The modes guide explains every mode.

Teach it your field's terms and the authors you cite

Every field has words a general speech model rarely hears: technical terms, acronyms, gene and compound names, and the surnames of the authors you cite most. YapToText ships no academic word list. You build your own on the Dictionaries page, where each entry pairs what the app hears, in Heard, with the spelling you want, in Replace with. Your dictionary works twice: it primes what Whisper listens for, then fixes what slips through.

Dictionary entry: An author's name, the way the microphone hears it and the way you cite it.

Heard: bore dew
Replace with: Bourdieu

What happens: Every bore dew in a dictation is written Bourdieu, and Bourdieu is handed to Whisper as a word to listen for. Whole word is on by default, so the entry never fires inside a longer word.

  • Use what it actually wrote. Heard has to match the speech model's spelling, not yours. Open the dictation in History, click Show pipeline details, and copy the raw transcript's version into Heard.
  • Add a name without typing it. Select an author's name spelled correctly anywhere on screen, tap Right Option, say add this to my dictionary, and tap again.
  • Catch the variants. If a term comes out several ways, press Generate variations on the row, and the model on your Mac maps likely mishearings to it.
  • One dictionary per project. Press New Dictionary and name it for the thesis or the grant. Each has its own switch, so turn one off when the project ends.
  • Mind the budget. The hint handed to Whisper holds about 200 characters, your dictionary terms first, and a Replace with longer than 40 characters is not used as a hint, though the replacement still runs. Put the terms you say most in the dictionary that is switched on.
YapToText's Dictionaries page: suggested fixes such as thursday to Thursday, and the General dictionary with iPhone and Wi-Fi.
Figure 2. The Dictionaries page: each entry pairs what the app hears with the spelling you want, and a dictionary per project can be switched off when the project ends.

The replacements run on the raw transcript and again after any AI cleanup, so cleanup cannot undo a fix. The Dictionaries guide covers Whole word, Case and order.

Leave citations to your reference manager

Citations are not speech. An in-text citation, a page range and a reference list entry are formatting, and saying "Smith 2019 page 14" gives you words and numbers, not your style guide. Dictate the claim, and let your reference manager or your keyboard write the citation.

Zotero's documentation says its word processor plugins work with Microsoft Word, LibreOffice and Google Docs, insert in-text citations and footnotes, and build a bibliography that updates as you cite. I have not tested them side by side with YapToText. YapToText only types text at your cursor, so the order is simple: dictate the sentence, then insert the citation with the plugin, or type it.

A placeholder keeps you talking when you do not want to stop and look up a source. On the Commands page, open Snippets, press Add Snippets, put cite marker under Say this, and put [CITE] under Inserts.

Say: With a snippet named cite marker saved on the Commands page.

Earlier studies found the same effect in older adults insert cite marker

What happens: The sentence is typed, followed by [CITE]. Snippets go in after any AI cleanup, so a rewrite cannot drop or move the marker. Later, search the draft for [CITE] and fill each one in.

While Require the word insert before a command is on, which is the default, a snippet only fires when you say insert in front of it, so an ordinary sentence about citing cannot set it off. The spoken punctuation guide covers that rule.

LaTeX: say the prose, type the markup

LaTeX is mostly markup: commands start with a backslash, and many take their arguments in curly braces. Every one of those is a symbol you would have to say, which is the same problem code has. Say the prose into the .tex file, and type the markup around it.

That is the idea in the guide to dictation for coding: say the prose, type the syntax. YapToText does have the names. Backslash is a one-word name, so you say insert backslash, and open brace and close brace are two-word names that fire at the end of a clause or before a capital letter or a number. Followed by a lowercase word, a two-word name stays as words, so dictating a command with its argument gets finicky fast. Type the commands, click inside the paragraph, and dictate the sentences.

Step 1: Type the structure

  • You: Type the section command, the environment or the figure, the way you always do, and leave the cursor where the prose goes.
  • Check: The cursor sits on its own line inside the section, not inside a command.

Step 2: Pick the mode for an editor

  • You: Tap Right Command, press 2 for Raw Transcription or 3 for Clean Up, and say the paragraph.
  • YapToText: With Auto mode on, Xcode, VS Code, Cursor, JetBrains IDEs, Sublime Text, Zed and the terminals on Auto's list lean toward Code Comment, which mostly takes dictations over 60 words and compresses them into short technical text. The number key overrides that for one dictation.
  • Check: In History, Show pipeline details names the mode that ran.

A LaTeX editor that is not on Auto's list, or one in your browser, gets no lean, so ordinary prose gets the careful pass that rewrites nothing. If you write LaTeX all day, turn Auto mode off and give your editor its own mode under Apps that use this mode. Equations are symbols from top to bottom, so type them. I have not tested LaTeX markup inside a snippet's Inserts box, where \n stands for a line break, so I type the markup too. If your editor plays the alert sound when the words land, turn off Intelligent Insert for it on the Dictation page, in the Delivery card.

Long writing sessions with no silence cutoff

A writing block for a thesis can run for hours, with long pauses while you think. Out of the box, YapToText does not mind. A pause to think does not end the dictation, and every take is kept.

  • No silence cutoff. Auto-stop after silence and Maximum length, on the Dictation page, Recording card, are both off by default. As a safety net, one recording stops itself after 30 minutes and still transcribes what you said.
  • Long takes in pieces. A dictation over 4 minutes is cut at quiet moments into pieces of about 2 minutes, transcribed one after another.
  • Pause on purpose. Press Space to pause while you check a source, and again to resume.
  • Apple Dictation, for comparison. Apple says its Dictation stops when it hears no speech for 30 seconds. The 30-second limit page explains it.
YapToText's Dictation settings page on a Mac: the Keys, Recording, Microphone and Delivery cards.
Figure 3. The Recording card on the Dictation page: Auto-stop after silence and Maximum length are Off by default, so a long pause to think does not end the take.

Keep every take until the thesis is done

  • You: In Settings, Advanced, History & audio, set Keep to Keep everything and Auto-delete old items to Never.
  • YapToText: Stops pruning. Out of the box it keeps your last 500 dictations and deletes anything older than 30 days.
  • Check: The two pickers read Keep everything and Never.

Export, at the bottom of the History page, saves the list you are looking at, after any search or filter, as one Markdown file, with each take under a heading that gives its mode and date. If the app or the Mac dies mid-sentence, the audio survives, and the next launch transcribes it. The Statistics page counts your words for each of the last 30 days, with streaks, which makes a quiet daily writing log; it counts what you said, not what survived the edit. The History guide covers search, playback and export.

Transcribe interviews for qualitative research

Qualitative research often starts with recorded interviews or focus groups, and YapToText transcribes a recording on your Mac with the same Whisper model it uses for dictation. Record only with consent and within what your ethics board approved. Pick Raw Transcription, drop the file, and quote only from the raw text and the audio.

  • Names first. Add participants' pseudonyms, place names and the field's jargon to a dictionary before you transcribe. Your dictionaries apply to file transcription as well as dictation.
  • The card. On the Utility page, pick Raw Transcription in the Transcribe a file card's mode menu, then drag the audio or video file onto it, or click Choose File. There is no length cap, and long files are handled in pieces.
  • Plain text. The result has no speaker labels and no timestamps, so you mark the turns by hand while you listen.
  • Quotes stay verbatim. Clean Up removes filler words and false starts, which makes readable notes and a changed quote. Check every quote against the recording.
  • What is kept. A file transcribed on the Utility card is not saved to History, and nothing is uploaded.

Whether that meets your ethics board's rules or your data management plan is for you and them to judge, and I make no claim about it. The guide to transcribing an interview on a Mac walks through recording, the steps on the card, adding speaker names and when a tool with speaker labels fits better.

What it cannot do

  • Write the paper. It types what you said. Raw Transcription never touches an AI model, and the argument, the analysis and the revisions are yours.
  • Cite or check sources. It spells an author's name the way your dictionary says. It does not know whether the source exists, says what you cite it for, or matches your style guide.
  • Write markup or equations. Its spoken commands are punctuation, emoji and snippets, so LaTeX commands, equations, headings and styles come from your keyboard and your writing app.
  • Label speakers or add timestamps. A transcribed interview is plain text.
  • Run anywhere but a Mac. It needs an Apple silicon Mac with macOS 14 or later, and nothing is claimed for Intel Macs.
  • Promise accuracy. I have not run an accuracy test on technical vocabulary, so read every dictated paragraph before it goes to your committee.

Dictation gets the draft down; the scholarship is still yours. YapToText is free on the Mac App Store, with no account and no paid tier.

The short version

  • Prose. Click into the draft, tap Right Command, say one paragraph, tap again.
  • Your own words. Turn Auto mode off, or press 2, for Raw Transcription with no AI cleanup.
  • Terms and authors. One dictionary per project, with the spelling the raw transcript shows in Heard.
  • Citations and LaTeX. Say a marker, then cite with your reference manager; type the markup and dictate the sentences.

Say the argument, type the apparatus, and keep every take until the thesis is done.

If your field's vocabulary or your editor gives YapToText trouble, open an issue on GitHub. Every one gets read.

Download YapToText on the Mac App StoreDownload for free. No account, nothing locked, macOS 14 or later on Apple silicon. Source on GitHub.

About the author. Ryleigh Newman is a developer whose hands make typing difficult and who dictates every day with YapToText, the app this site describes. Facts about other apps come from their own sites, with the date they were read. More about the project.

Sources

Questions

Can I write a thesis using dictation?

Yes, for the prose. Dictate one paragraph or one argument at a time, leave a marker where each citation goes, and do the precise writing in a separate editing pass. Citations, equations and markup are faster to type or insert.

What is the best dictation software for academic writing on a Mac?

YapToText, which I built, is free, transcribes on your Mac, has no silence timeout by default and types into Word, Google Docs or a .tex file; weigh that I made it. Apple Dictation is free and built in, but it stops after 30 seconds without speech.

Can dictation handle technical terms and author names?

With help. Add each term or name on YapToText's Dictionaries page, with the spelling the speech model produced under Heard and yours under Replace with. The entries also prime the speech model, so the word is more likely to come out right the first time.

Can I dictate in LaTeX?

Dictate the prose and type the markup. LaTeX commands start with a backslash and take arguments in curly braces, and saying each symbol is slow, so type the structure and dictate the sentences inside it.

Does dictation work with Zotero?

Zotero's docs say its plugins insert citations in Word, LibreOffice and Google Docs. YapToText only types text at your cursor, so dictate the sentence, then insert the citation with the plugin. I have not tested the two side by side.

Is dictation allowed under my university's AI rules?

Check the rules of your institution, and of any journal or funder you submit to; I make no claims about them. In YapToText, Raw Transcription types your exact words with no AI cleanup, while the other modes use an on-device AI model.