Quick Edit: edit selected text by voice

To edit text by voice on a Mac with YapToText, select the text in any app, hold the Quick Edit key (Right Option by default), say what you want changed, and let go. The rewrite replaces your selection in place, and the whole thing runs on your Mac.

I dictate everything because my hands make typing difficult, and fixing a sentence I had already written used to mean going back to the keyboard. Quick Edit is how I avoid that. I select the words and say the change, the same way I would lean over to a person at the keyboard and tell them the edit out loud.

How it works

Quick Edit rewrites text that already exists. It is on by default.

1. Select the text you want changed, in any app.
2. Hold the Quick Edit key, Right Option by default, and say the change.
3. Release the key. The selection is replaced with the result.

It works on anything you can select: an email you are drafting, a comment in your code, a message you have not sent yet. What you say is treated as an instruction, not as text to insert, so nothing you say is typed literally.

What to say

Talk to it the way you would talk to an editor. Short, plain instructions work best.

• Make this shorter
• Fix the grammar
• Make it sound more formal
• Turn it into bullet points
• Translate it to Spanish
• Capitalize this word
• Make this past tense

One instruction is special. Select a word or short phrase and say add this to my dictionary, and the app saves it as a dictionary entry: from then on, when the recognizer hears that word, it is written with the exact spelling and capitalization you selected. The card says Added with the word in quotes, and the entry is in place right away without waiting on a rewrite. For a single word, the app then asks the cleanup model which spellings a recognizer is likely to produce for it and maps those to the same word, so a name still comes out right when it is misheard. If the word is already in a dictionary, the app tells you instead of adding a duplicate.

What you see on screen

A small card appears near the bottom of the screen the moment you press the key. It never takes focus away from the text you are editing.

• While the key is down it says Listening for your edit and shows a live waveform. Until you speak it says Say what to change, then release the key. As you speak, the words it hears appear in quotes underneath.
• When you let go, the waveform winds into a spinning ring. The card says Got it while your words are transcribed, then Applying your edit, with the instruction it heard in quotes and Press Esc to cancel underneath.
• When the text has been replaced, the ring turns green and the card says Done, then fades away on its own. If the edit could not be applied, the ring turns red and the card says Couldn't apply that edit.

Cancel with Esc

While the card says Applying your edit, press Esc. One press is enough. The card changes to Cancelled - nothing was changed, and your selection is left exactly as it was. The result is thrown away before anything touches your document. Pressing the Quick Edit key again while an edit is applying also cancels it.

If you release the key before saying anything, the card says Didn't catch an instruction and nothing changes.

Which model does the work

Two models take part, and both run on your Mac. The dictation model, Whisper Large v3 Turbo, turns your spoken instruction into words. The cleanup model then applies that instruction to the selected text. Out of the box Quick Edit uses the same cleanup model as dictation, which is the bundled Phi-3.5 Mini unless you have changed it on the AI Models page.

The Quick Edit page has its own Model for edits picker. Leave it on Same as dictation, or choose any cleanup model you have downloaded. Only models already on disk are offered, and the page links to AI Models if you want to download more. Picking a separate model here leaves dictation cleanup on its own model.

Nothing is uploaded. The selection, your instruction, and the rewrite never leave the machine, and Quick Edit instructions are kept out of History. The privacy page has the full picture.

Change the key or how it fires

On the Quick Edit page you can bind a different key and choose how it fires:

• Hold to talk, the default: hold the key while you speak, release to apply.
• Tap to start / stop: tap once to start listening, tap again to apply.
• Off: the key does nothing.

Any key can be a trigger, not just a modifier. Caps Lock is a hardware toggle with no held state, so on Caps Lock hold to talk behaves as tap to start and stop. With Right Option bound, Option shortcuts and accented characters keep working normally.

When nothing happens

• Select some text first. If nothing is selected, the card tells you so. In hold mode, highlight the words and hold the key again. In tap mode the card stays up for about twelve seconds and starts listening by itself the moment you highlight something.
• Can't read the selection. Secure Input is on, which usually means a password field. Click into normal text and try again.
• Accessibility permission. Reading your selection and typing the rewrite back both need the Accessibility permission, the same one that lets the app type into other apps. Without it, selected text cannot be read.

Still stuck? Report it on the GitHub issues page.

Questions

All help topics · YapToText home