Mac Dictation Not Working in Figma? 9 Fixes That Actually Work

Short answer: Figma draws its own canvas instead of using normal text fields, so dictation only lands if a text layer is already in edit mode. If a layer is merely selected, your spoken words are read as single-key tool shortcuts. Double-click into the layer first, then check microphone permission.

Dictation in Figma fails in a way that feels personal. In Mail, in Slack, in a browser form, you press your shortcut and words appear. In Figma you press the same shortcut, the microphone indicator shows up, you speak a sentence, and either nothing happens at all or something far stranger happens: the canvas fills with rectangles, your zoom level jumps, a frame appears out of nowhere, and the text you dictated is nowhere on screen.

That second outcome is the tell. It is not a broken microphone and it is not a Figma bug. It is the predictable result of a design tool that treats every keystroke as a potential command, receiving forty words of dictated text while no text field has focus.

Below is what is actually happening, followed by nine fixes in the order that resolves the most cases fastest. Work down the list and stop when your text lands.

Why Figma is different from every other app on your Mac

Almost every app you dictate into uses standard text controls. A Mail compose window, a Notes document, a search box in Safari: these are text fields the operating system understands. macOS Dictation transcribes your speech and inserts the result into whichever text field currently has focus. The app barely participates.

Figma does not work that way. The design surface is a canvas that Figma draws itself, pixel by pixel. There is no operating-system text field sitting under your text layer. When you type on the canvas, Figma is intercepting keystrokes and deciding what each one means based on the current state of the editor. That gives Figma the responsiveness a design tool needs. It also means dictation has nowhere obvious to go.

Crucially, Figma runs in two very different states, and single keystrokes mean opposite things in each:

• Selection state. A layer is selected but you are not editing its content. Single letters are tool shortcuts: V switches to the move tool, T to text, R to rectangle, F to frame, O to ellipse. This is the state most people are in when they start dictating.
• Text edit mode. You have double-clicked into a text layer and a blinking cursor is inside it. Now letters are letters, and dictated text has somewhere to land.

Dictate a sentence in selection state and Figma receives it as a rapid stream of unrelated commands. That is the rectangle storm. Dictate the same sentence in text edit mode and it usually behaves like any other app.

The other surfaces in a Figma file are a mix. Comments, layer names in the layers panel, the file name, the search field, and most fields in plugin panels are real text inputs and behave normally. The canvas is the odd one out. Knowing which surface you are aiming at is half the diagnosis.

Start here: the 30-second split test

Before you change any setting, find out whether the problem is macOS Dictation itself or the way Figma handles it.

1. Open TextEdit and create a new blank document.
2. Press your dictation shortcut and say a short sentence.
3. Watch what happens.

If nothing appears in TextEdit, dictation is broken at the system level and Figma is innocent. Skip to fix 3 and work forward.

If the sentence appears cleanly in TextEdit, macOS Dictation is fine. The failure is in how Figma receives it, so start at fix 1.

There is a useful third test. Open a Figma file, press C to grab the comment tool, click anywhere on the canvas to open a comment box, and dictate into that. Comment boxes are ordinary text inputs. If dictation works in a comment but not on a text layer, you have confirmed the canvas is the problem and fix 1 is almost certainly your answer.

Fix 1: Enter text edit mode before you speak

This resolves more Figma dictation complaints than every other fix combined.

Do not select a text layer and start talking. Get a cursor inside it first:

• Double-click the text layer on the canvas, or
• select it once and press Enter, or
• press T, drag out a new text box, and start with the cursor already inside.

You want a blinking insertion point visible inside the text before the microphone opens. If you can see the caret, dictation has a destination.

A related trap: clicking once on a text layer inside a frame often selects the frame, not the text. Figma nests selection, so the first click may land a level or two above what you meant. Double-clicking repeatedly drills down. Check the layers panel or the selection outline to confirm you are actually inside the text layer and not on its parent.

If you already caused a shortcut storm

Press Cmd+Z repeatedly. Every stray shape and every accidental tool switch is a separate undo step, so you may need a dozen presses. Undo in Figma is reliable, and a shortcut storm looks far more destructive than it is. Resist the urge to manually delete things first; undo will unwind it more cleanly than you will.

If you were working in a shared file and the storm reached collaborators, Figma's version history lets you restore the file to a point before the mess. Open the file menu and look for version history, then restore or duplicate from an earlier autosaved state.

Fix 2: Confirm the dictation shortcut is actually firing

If no microphone indicator appears at all when you press the shortcut, dictation never started, so nothing downstream matters.

Open System Settings, go to Keyboard, then Dictation, and confirm the toggle is on. Check the Shortcut setting while you are there. If it is set to the Globe key, open the Keyboard pane's main screen and look at the Press 🌐 key to dropdown. On many Macs this is set to Change Input Source or Show Emoji & Symbols, which means your Globe press is doing something else entirely and dictation never gets a chance.

Set the shortcut to something unambiguous, such as pressing Control twice, and test again in TextEdit. If the indicator now appears in TextEdit but not while Figma is frontmost, a Figma keybinding may be swallowing the shortcut. Figma binds a lot of keys; picking a modifier combination Figma does not use avoids the collision entirely.

Fix 3: Repair macOS Dictation at the system level

If the TextEdit test failed, work through these in order:

• In System Settings, Keyboard, Dictation, toggle dictation off, wait a few seconds, then back on. This re-registers the service and clears a surprising number of stuck states.
• Check the Microphone Source in the same panel. If it points at a device you have unplugged, dictation is listening to nothing. Set it to Automatic or pick the mic you are actually using.
• Confirm the language listed matches the language you speak. See fix 9.
• Restart the Mac. Dictation depends on background services that occasionally need a clean start, particularly after a macOS update.

If dictation was working before a recent system update and stopped afterwards, that specific pattern has its own set of causes, covered in our guide to Mac dictation not working after a macOS update.

Fix 4: Give microphone permission to the right app

This is where the desktop app and the browser diverge, and where people spend the most wasted time.

Open System Settings, then Privacy & Security, then Microphone. Look at the list. If you use the Figma desktop app, Figma should be listed and enabled. If you use Figma in a browser tab, it is your browser that needs the entry, not Figma.

Two clarifications that save a lot of confusion:

• macOS Dictation itself does not require the frontmost app to have microphone permission. Dictation captures audio at the system level and delivers text as keyboard input. If dictation works in TextEdit, your microphone hardware and system permission are fine.
• App-level microphone permission matters for anything the app records itself, and for any third-party dictation tool you run. If you use a separate dictation app, it needs its own entry under Microphone, and often under Accessibility or Input Monitoring as well so it can send text into other applications.

After granting a permission, quit and reopen the app. macOS often does not apply a newly granted microphone or accessibility permission to an already-running process.

Fix 5: In the browser, check site permission and the per-profile trap

If you use Figma at figma.com rather than the desktop app, there is a second permission layer.

Click the icon at the left edge of the address bar and confirm the microphone is set to Allow for figma.com. In Chrome you can also review this at chrome://settings/content/microphone. Again, this only affects features that record audio in the page, not macOS Dictation, but it is worth clearing so you can rule it out.

The trap that catches people: Chrome permissions are per-profile. If you have a work profile and a personal profile, granting access in one does nothing for the other. Confirm which profile the Figma tab is actually open in, then grant permission there.

Browser dictation problems have a broader set of causes than Figma alone, including extension conflicts and managed policy. Our walkthrough of Mac dictation not working in Chrome covers those in detail, and everything in it applies to a Figma tab.

Fix 6: Stop Escape and click-away from discarding what you dictated

Some people find their dictated text appears on the canvas and then vanishes. That is usually not dictation failing. It is how you exited text edit mode.

In Figma, Escape leaves text edit mode. Many dictation workflows involve pressing Escape reflexively to close the dictation indicator, and if the dictated text has not been committed yet, that keystroke can take it with it. Instead of Escape, click on empty canvas to commit the text, or press Cmd+Enter, then close the dictation session.

The same applies to clicking away mid-dictation. If dictation is still processing when you change focus, the insertion has no valid target by the time it completes. Let the dictation session finish and the text settle before you touch anything else.

Fix 7: Plugin panels, Dev Mode, and embedded frames

Figma plugins run in sandboxed frames that are separate from the main editor. So do some embedded views. Text fields inside them are usually real inputs, which is good, but focus behaves differently:

• Click directly into the plugin's text field and confirm you see a caret before you start speaking. Clicking the panel background is not enough.
• Some plugin panels re-render as they load data, which discards focus mid-sentence. Wait for the panel to finish loading before dictating.
• If a plugin field consistently rejects dictated text, dictate into the canvas or a comment first, then copy and paste into the plugin. This is a workaround rather than a fix, but it is reliable.

Dev Mode and prototype view are largely read-only surfaces. If you are trying to dictate into a Dev Mode panel and nothing lands, check that the field you are aiming at is actually editable in that mode.

Fix 8: Free the microphone from calls and recorders

Design work involves a lot of screen sharing, and the microphone is not always shareable. If a video call, a screen recorder, or a voice channel is holding exclusive access to the input device, dictation may open and capture silence.

Quit the other application rather than muting it. Muting inside a call app often keeps the device claimed. Fully quit video call clients, recording tools, and voice chat apps, then test dictation again in TextEdit before returning to Figma.

Bluetooth headsets add a wrinkle of their own. When a headset switches into its bidirectional call mode, audio quality drops sharply and some transcription paths handle the resulting input poorly. If you dictate with wireless earbuds, try the same sentence with the built-in Mac microphone as a comparison.

Fix 9: Fix the language mismatch that turns clean audio into nonsense

This one is easy to misdiagnose because dictation appears to be working perfectly. Words arrive promptly and confidently. They are just the wrong words.

If the configured dictation language does not match what you are speaking, the system does its best to map your audio onto the vocabulary it was told to expect. You get fluent, well-punctuated nonsense. Open System Settings, Keyboard, Dictation, and check the Languages list. If you have multiple languages enabled, confirm the active one is the one you are speaking, and remove any you do not need.

A softer version of the same problem affects design vocabulary. Terms like auto layout, variant, constraint, tokens, and component names specific to your design system are ordinary words in general English, so a general-purpose transcriber has no reason to prefer your meaning. Hex codes and spacing values suffer worst of all.

The workflow that sidesteps most of this

Every fix above addresses a symptom of the same underlying design: macOS Dictation is a mode you enter, and while you are in it, focus is fragile and stray keystrokes are dangerous.

Voice Keyboard Pro takes a different shape. It lives in the menu bar. You hold a hotkey, speak, and release, and the text appears at your cursor in whatever app is frontmost. There is no dictation window to dismiss, no session to end with Escape, and no ambiguity about when capture starts and stops, because you are holding the key the entire time you are speaking.

For Figma specifically, that hold-to-talk shape matters more than it does elsewhere. Because the capture window is bounded by the key you are physically holding, you can double-click into a text layer, confirm the caret, hold the key, speak, and release, with no intermediate keystroke that could drop you back into selection state. The text lands where the cursor is.

The feature that pays off fastest for designers is Smart Vocabulary, a personal dictionary with replacement rules. Teach it your design system's component names once and they come out spelled and cased the way your team writes them, every time. The same applies to product names, client names, and the acronyms that are unique to your organisation. Instead of correcting the same six words on every text layer, you fix them once.

If you also run design reviews and critiques, Meeting Mode captures the session with speaker detection and produces AI notes, so feedback that would otherwise live in someone's memory ends up as text you can turn into comments and tickets. Calendar meeting detection means it can offer to start when the review actually begins.

On privacy, worth stating plainly: our server stores only operational pings. No audio and no transcript content is retained.

What voice is genuinely bad at in Figma

Honesty is more useful here than enthusiasm. Dictation is excellent for some Figma work and pointless for the rest.

Voice is good for: body copy and paragraph text in mockups, filling realistic content into a design instead of placeholder text, writing comments and critique feedback, naming layers and frames descriptively, drafting handoff notes, and writing the annotation text that explains a flow.

Voice is bad for: hex codes, spacing and sizing values, anything numeric that needs to be exact, precise positioning, and short single-word labels where saying it takes longer than typing it. It is also the wrong tool for navigating the canvas. Zoom, pan, and selection are keyboard and trackpad work.

The realistic gain is a straightforward speed difference. Most adults type around 40 words per minute and speak comfortably at 130 to 150. Even strong typists at 80 to 100 words per minute are slower than their own speech. On a mockup that needs three paragraphs of believable copy, that gap is the difference between a two-minute task and a thirty-second one.

Dictation in Figma is not about talking to your design tool. It is about getting real words into text layers faster than you can type placeholder ones.

Quick reference

• Rectangles appear when you speak: you are in selection state. Double-click into the text layer first.
• Nothing happens anywhere, including TextEdit: system-level problem. Toggle dictation off and on, check microphone source, restart.
• No microphone indicator at all: the shortcut is not firing. Check the Globe key setting and pick a non-conflicting shortcut.
• Works in comments, fails on canvas: confirmed canvas focus issue. Fix 1.
• Text appears then disappears: Escape or a click-away dropped it before it committed.
• Works in one browser profile, not another: Chrome microphone permission is per-profile.
• Fluent output, wrong words: language mismatch in the Dictation settings.
• Silence captured during a call: another app is holding the microphone. Quit it, do not just mute.

If your work spans the wider Figma ecosystem, our guides to dictating in Figma and dictating in FigJam cover the day-to-day workflow rather than the failure modes, and voice to text for UX researchers is worth a look if you write up sessions and interview notes alongside your design files.

Most Figma dictation problems are one of two things: a text layer that was selected rather than opened for editing, or a microphone permission granted to the wrong app. Check those two first and you will resolve the majority of cases before you get to anything complicated.