Dictation modes: the same words, shaped for where they are going
By Ryleigh Newman · Published · Updated
Dictation modes decide what YapToText does with your words after Whisper hears them on your Mac: Raw Transcription types them exactly, and Clean Up, Note, Email, Message and Code Comment rewrite them for that kind of writing while keeping your words. Auto mode, on by default, picks one for you each time.
The modes live on the Modes page. Every one is editable, built-ins included. You can write your own in plain sentences, and any app can have its own mode. Everything runs on your Mac, so a mode never sends your words anywhere.
This app is completely free. You don't have to make an account. There are no subscriptions, no difficult downloads, and nothing is locked. Made for Apple silicon Macs. Source on GitHub.
This guide in other languages: العربية, Čeština, Dansk, Deutsch, Ελληνικά, Español, Suomi, Français, עברית, हिन्दी, Magyar, Bahasa Indonesia, Italiano, 日本語, 한국어, Norsk, Nederlands, Polski, Português, Română, Русский, Slovenčina, Svenska, ไทย, Türkçe, Українська, Tiếng Việt, 简体中文, 繁體中文
On this page: 9 sections
What each mode does
Six modes ship with the app, plus Auto:
- Auto. Reads each dictation and picks the treatment by itself. It has its own guide: Auto mode.
- Raw Transcription. Fast, literal speech to text with basic punctuation, and no AI. It is the selected mode whenever Auto is off, until you pick another.
- Clean Up. Removes filler words and false starts, fixes grammar, capitalization and punctuation, and repairs words the microphone clearly misheard. It keeps your wording, tone and register, blunt or profane language included, never deletes, shortens or summarizes a sentence, writes spoken web and email addresses in proper form, and turns spoken numbers into digits without adding units. The app calls it the recommended default.
- Note. Short headings and bullets, with action items pulled into a list, and nothing added that you did not say.
- Email. An email body, never a Subject line. It adds a greeting only if you named someone, and signs with the name in Settings, Advanced, Your name, never with a placeholder like [Your Name].
- Message. Light cleanup for chat: filler gone, obvious grammar fixed, the casual tone kept.
- Code Comment. Concise, technical phrasing for code comments and commit messages, with identifiers and casing kept exactly as spoken and no quotes or code fences.
Every mode but Raw Transcription hands the transcript to the cleanup model, Phi-3.5 Mini Instruct by default, with its own instructions. If the result reads like an assistant's reply, grows well past what you said or repeats the reference text, the app throws it away and types the raw transcript instead. Raw Transcription never touches an AI model. Every other mode keeps your words and fixes the rest.
Pick a mode, or switch in the middle
There are four ways to pick a mode:
- From the menu bar. Click the capybara and use the Mode picker. Picking a mode there turns Auto off, and Auto leaves the picker until you switch it back on, on the Home page or in the Auto row on the Modes page.
- While you talk. Press 1 to 9, or click a mode chip on the expanded panel. With Auto on, the pick lasts for this dictation. With Auto off, it becomes your selected mode.
- From the Modes page. Right-click a mode and choose Use This Mode. The current one carries a Current tag.
- With a shortcut. Cycle mode in Settings, General, Shortcuts steps through the list, and any mode can have its own Activation shortcut that starts a dictation in that mode.
The numbers follow the order of the Modes list, so drag a mode up or down and its number changes with it. A mode with Show in the quick switcher turned off gets no number. The number on a mode is the key that picks it. Every key is in YapToText's keys.
Edit a mode, or make your own
Click any mode on the Modes page to open its editor. Its cards are Basics (name, icon, description, Show in the quick switcher, Activation shortcut), Apps that use this mode, Speech model with an On battery choice, Recording & language, AI cleanup, Dictionaries, Where the text goes and Try it out.
Step 1: Start from a mode that is close
- You: Right-click a mode and choose Duplicate, or press the plus button and choose New Mode.
- YapToText: Opens the new mode's editor.
- Check: The new mode appears in the list on the Modes page.
Step 2: Write its instructions
- You: In the AI cleanup card, keep Rewrite the transcript with AI on and write what you want in the Instructions box, or press a Start from button to begin from a built-in mode's instructions.
- YapToText: Saves as you type. The cleanup model follows these instructions for every dictation in this mode.
- Check: The instructions say what to keep as well as what to change.
Mode instructions: An example for short replies in a chat app.
Turn my dictation into a short, friendly reply. Keep my words and my meaning. Do not add anything I did not say.What happens: The cleanup model rewrites each dictation in this mode toward that, on your Mac.
Step 3: Try it out
- You: In the Try it out card, type a sample and press Preview with AI.
- YapToText: Runs the mode on the sample, without dictating.
- Check: The Result reads the way you want. If not, edit the instructions and preview again.
Under AI cleanup you also pick the Cleanup model, and three switches set what context the model may see. Built-in modes have Reset to Default on their right-click menu, Restore Built-in Modes is under the plus button, and the Undo button steps back the last mode change. Export Mode and Import Mode move a mode as a file. A mode is just instructions you can read and rewrite.
Give an app its own mode
Any app can have its own mode, so dictation there uses it without you picking it. Put Mail on Email, Messages on Message and your editor on Code Comment.
- You: In the mode's editor, under Apps that use this mode, press Add app and pick a running app, or Choose App for any other. Settings, Advanced, Per-app modes edits the same list.
- YapToText: Switches to that mode whenever the app is in front when a dictation starts. This needs no Accessibility permission.
- Check: With Auto mode off, a dictation in that app comes out shaped by its mode.
While Auto mode is on, Auto decides how each dictation is written, so per-app modes shape the writing only when Auto is off. If you dictate into many apps, the guide to dictating into any app on a Mac goes app by app.
Run a mode on text you already have
Three places run a mode on text that is already written:
- Regenerate in the menu bar. Click the capybara, then Regenerate: Redo it as runs your last dictation through any mode, Rewrite it as rewrites the text you have selected in any mode, and a one-shot edit applies an AI action.
- History. Every entry keeps its raw transcript, so Regenerate with AI reruns an old dictation as a different mode. The History guide has the details.
- Utility. The Transform text card takes pasted text and reformats it through any mode, like turning notes into an email. The Utility guide covers the rest of that page.
Rewriting selected text needs Accessibility, the same permission that lets the app type. A mode works on any text, including words you dictated last week.
Limits
- A cleanup model. Every mode but Raw Transcription needs one. The Homebrew build ships without models, so download Phi-3.5 Mini Instruct, 2.3 GB, on the AI Models page before relying on an AI mode.
- Post-transcription analysis. When this master switch on the Home page is off, no mode runs: every dictation is typed exactly as spoken, and the Modes page says so.
- A small model. The cleanup model runs on your Mac, so preview a new mode in Try it out before you rely on it.
- Mid-dictation switches. A new mode's cleanup applies to the dictation in progress, but its speech model and language apply from the next one.
- Nine numbers. Number keys reach the first nine modes in the switcher.
If a mode does something odd
- See which mode ran. In History, click Show pipeline details on that dictation. The Mode line names it, and Auto detected says what Auto picked.
- Check the master switch. On the Home page, in the After transcription card, Post-transcription analysis must be on.
- Preview the same words. Paste what you said into Try it out in the mode's editor and compare.
- Fix the instructions. Say what to keep as well as what to change, or use Reset to Default on a built-in mode.
- Try another model. Pick a different Cleanup model in the mode's editor, or use Raw Transcription for that kind of writing.
- Tell me. Open a GitHub issue with the mode, what you said, and what came out.
The short version
- Raw Transcription. It types exactly what you said, with no AI.
- Clean Up and friends. They tidy your words for an email, a note, a chat or a code comment.
- Auto. It is on by default and picks for you.
Pick the mode by where the words are going, or let Auto pick. If you are coming from superwhisper, the superwhisper alternative page compares the two. If a mode surprises you, open an issue on GitHub. Every one gets read.
This app is completely free. You don't have to make an account. There are no subscriptions, no difficult downloads, and nothing is locked. Made for Apple silicon Macs. Source on GitHub.
Related
Questions
Which YapToText mode should I use?
Auto, which is on by default, picks for you. If you pick one yourself, Clean Up drops the ums and fixes grammar without changing your meaning, and Raw Transcription types exactly what you said.
Does Raw Transcription use AI?
No. It is fast, literal speech to text with basic punctuation: Whisper hears you, and nothing rewrites the words.
Can I make my own dictation mode?
Yes. Duplicate a mode that is close, or choose New Mode under the plus button on the Modes page, then write its instructions in plain sentences and test them in Try it out.
Why does Email mode not write a subject line?
Its instructions write the email body only. Every built-in mode is editable, so you can change that.
Can each app use a different mode?
Yes. Add the app under Apps that use this mode in the mode's editor, or in Settings, Advanced, Per-app modes. While Auto mode is on, Auto decides the writing instead.
Do modes send my text anywhere?
No. The speech model and the cleanup model both run on your Mac, and your words and instructions stay there.