Mac Dictation: The Complete 2026 Guide

By the Viora team · Last updated April 28, 2026.

Most people who tried dictation on their Mac three or four years ago tried it once, dictated a sentence with the word "comma" in it, watched the word "comma" appear instead of an actual comma, and never opened the feature again. That's roughly the spot Mac dictation was in until macOS Sonoma. Things have moved since then.

This page is the version we wish we'd had when we started building Viora. It walks through the built-in macOS dictation feature step by step, says plainly when it's enough and when it isn't, and gives an honest read on the four third-party voice-to-text apps actually worth your time on a Mac in 2026 — Superwhisper, Wispr Flow, Typeless, and our own product, Viora.

We're the Viora team writing on our own blog, so the obvious disclaimer applies. We try to handle our competitors the way we'd want a competitor to write about us. Name the things they do better than us, link to their site, trust the reader.

What's actually different about Mac dictation in 2026

Two things changed in the last couple of years and they made the built-in feature go from novelty to actually useful for short input.

First, Mac dictation now runs on-device for most major languages. Once macOS downloads the language pack the first time, your audio doesn't leave your Mac. That covers English, French, German, Spanish, Mandarin, Cantonese, Japanese, and a handful of others — Apple's official dictation support page is the reference for the full list. Less common languages still go through Apple's servers, but for the languages most people actually dictate in, you can disconnect from Wi-Fi and it still works.

Second, Apple shipped a noticeably better neural model for common nouns and short phrases. Names of people you've messaged, addresses you've typed, technical terms with local context — all materially better than they used to be.

What hasn't changed is the part that decides whether you keep using it.

You still have to say punctuation out loud. "Send the report by Friday period new line let me know if you need anything else period" is how you actually have to dictate a two-sentence email. You can dictate "ok thanks" into Messages and it'll feel fast. You can't dictate a paragraph without losing patience.

There's still no transcript editor. Text appears directly in whatever field you're typing into. If you started a sentence over because you changed your mind halfway, both versions are now in your draft. You fix it the same way you'd fix any typo.

There's still no public dictionary editor. Names, products, internal jargon — there's no way to teach the built-in feature to spell them. macOS adapts implicitly over time, but you can't open a list and add the word Composio to it.

These three gaps are exactly what the third-party apps exist to fill. We'll come back to them.

How to enable dictation on your Mac

Settings moved between macOS versions, so the click path depends on which one you're on. As of macOS Sonoma 14 and Sequoia 15, here it is.

1. Open System Settings (⌘ + Space, type "System Settings").
2. In the sidebar, scroll to Keyboard.
3. Find the Dictation section about halfway down the right pane.
4. Toggle Dictation on. macOS will ask for microphone permission. Grant it.

Done. The feature now works in any text field on your Mac.

But you almost certainly don't want the default shortcut. The default is "press the dictation key on a newer Apple keyboard" or "press fn twice in quick succession." Neither feels right. The first one only works on hardware that has the dedicated key. The second is a tap-twice combo that gets uncomfortable somewhere around the third time you do it.

Click Shortcut in the same panel and pick a hold-to-talk key. Anything you can press and hold without it conflicting with something else:

• Right ⌘ held down
• fn held down (works on every Mac keyboard)
• Right ⌥
• Right ⌃

The mental model becomes "while I'm holding this, I'm dictating." Press, talk, release. Same shape as a walkie-talkie. We've watched dozens of people set this up. The ones who pick a hold-to-talk key keep using dictation a week later. The ones who keep the tap-twice default don't.

Then click Languages and add only the ones you actually speak. Auto-detection is good but it's not magic. If you have French enabled but never speak French, dictation will occasionally route a borderline-sounding sentence through French and produce garbage. Two or three languages max.

For the very first test, put your cursor in a Notes window, hold your shortcut, and say "this is a test of dictation on my mac." You should see the words appear, with proper capitalization on "Mac."

If nothing happens, the most common cause is microphone permission. Check System Settings → Privacy & Security → Microphone. macOS Dictation needs its own slot in that list, and the app you're focused on might also need permission depending on how it's sandboxed.

If words appear but in the wrong language, go back to Languages and remove the ones you don't actually use.

If it works for a few seconds and then cuts off mid-sentence, that's the feature, not a bug. Built-in dictation auto-stops after a few seconds of silence. There's no setting to change this. It's one of the bigger reasons people end up installing a third-party tool.

The dictation commands you actually need

You can ignore most of the dictation commands list Apple ships. For 95% of dictation, six commands cover what you need:

• "comma" → ,
• "period" → .
• "question mark" → ?
• "exclamation point" → !
• "new line" → cursor goes to next line
• "new paragraph" → blank line, then next paragraph

A handful more that show up enough to be worth remembering: "open quote" / "close quote" for ", "colon" for : (yes, you have to say it), "slash" for /, "at sign" for @ (says itself in some contexts but not all), "caps on" / "caps off" to toggle capitalization, and "stop dictation" to end the session without releasing the key.

If you find yourself saying "period new line let me know if" four times inside a single email, that's the feature working as designed. It's also the moment you realize built-in dictation was built for short input, not for writing.

When the built-in feature isn't enough

We watch for three signals when someone asks whether they should bother installing a third-party tool.

You spend more time editing than dictating. If your raw transcript needs cleanup before it can be sent, you're not actually saving time. You're just shifting work from typing to editing. Modern third-party tools run a small language model on top of the transcript that handles punctuation, capitalization, and filler-word removal automatically. You stop saying "comma" out loud.

You keep correcting the same names. The built-in feature has no public dictionary. If you have a colleague named Nguyen, an internal product nobody outside your company has heard of, or technical jargon that only makes sense in your domain, the built-in dictation will get them wrong forever. Personal dictionaries are one of the biggest differences between built-in and third-party.

You want voice for more than just text. This is where the field is moving. Voice as input to a writing surface is one thing. Voice as input to a calendar lookup, a search, a tool call — that's a different category. Only one of the four apps below does this today; the other three are still pure dictation tools, and that might be exactly what you want.

If none of these signals are showing up for you, stop reading here. The built-in feature is fine for what it is, free, and it ships with your Mac.

The four third-party apps actually worth knowing in 2026

We watch this space pretty closely. There are easily a dozen voice-to-text apps marketed for Mac, and most of them are wrappers around the same underlying transcription model with marginally different UX. These four are the ones with enough scale, signal, or unique positioning that we think they're worth your evaluation time.

We'll go through each one as honestly as we can — what it does well, what it doesn't, and roughly who it's for. Pricing and platform support were verified against each vendor's published page on April 28, 2026.

Superwhisper — the offline-first option

Superwhisper is the reference point on Mac for "no audio ever leaves my machine." It bundles a deep library of local Whisper variants — Whisper Ultra V3 Turbo, Parakeet via WhisperKit, several smaller variants — and lets you bring your own LLM API key (Claude, GPT, Llama, others) for the post-processing step. The whole pipeline runs on your Mac if you choose the local options. Their privacy page is unambiguous about audio never leaving the device.

Pricing is the cleanest part of the category. Pro is $8.49/month or $84.99/year, but the headline number is the $249.99 lifetime license. Pay once, use it on every Mac you'll ever own. In a category where everyone else is subscription-only, that's the kind of pricing that doesn't need marketing.

What it doesn't do is agent mode. The "Super" mode pipes a transcript into the LLM of your choice, which is closer to "ask GPT a question with my voice" than to a tool-using agent. If your sentence is "schedule a 30-minute focus block tomorrow," Superwhisper will hand back a draft of that sentence, not the actual calendar event. We've put a side-by-side with Viora up if you want the longer comparison.

Pick Superwhisper if offline matters, you'd rather pay once than subscribe, or you specifically want to choose your own local Whisper model.

Wispr Flow — the cross-platform incumbent

Wispr Flow is the most established product in the category by a meaningful margin. They run on Mac, Windows, iPhone, and Android, with a synced personal dictionary that follows you across all four. They publish a sub-700ms p99 end-to-end latency figure, which is fast enough that you stop noticing the round trip — the Baseten case study goes into how they hit it.

Pricing is $12 per user per month billed annually for unlimited Pro use. The free tier is 2,000 words/week on Mac and Windows, 1,000 words/week on iPhone. They've also done the compliance work — HIPAA-ready on every plan, SOC 2 Type II at the Enterprise tier — which matters if you're in a regulated industry.

What they're known for in our circles is that the editing layer is good. AI Auto Edits with four intensity levels. Personal dictionary that learns your specific words. Tone that adjusts based on which app you're in. If most of your dictation goes into Notion, Gmail, and Slack, this is the tool that's most aggressively engineered around those exact workflows.

What it doesn't do is offline (per their privacy page, "Transcription always happens in the cloud") and it doesn't have agent capabilities. Picking Wispr Flow over Superwhisper is mostly a privacy and pricing call.

Pick Wispr Flow if your day moves between platforms, you need 100+ language coverage, or HIPAA compliance is a hard requirement. Our Viora vs Wispr Flow comparison has the deeper read.

Typeless — the editor that's the editor

Typeless is the newest of the four to scale. Three #1 Product Hunt launches in five months — desktop in November 2025, iOS in December 2025, Android #3 in January 2026 — which is the kind of cadence that says the team is shipping aggressively.

Their angle is the cleanup layer. The marketing line is "AI Voice Dictation That's Actually Intelligent" and what they mean is that the editing model does more than punctuation. It removes filler ("um", "you know"), restructures lists when you ramble through them, swaps tone based on which app you're in, and lets you "Ask Anything" about selected text on the fly. If your day is mostly long-form writing — email, docs, threads — this is probably the tool that does the most heavy lifting on the cleanup side.

Pricing is $12/seat/month billed annually, or $30/month month-to-month. Free tier is 8,000 words/week, the most generous of the four. They run on Mac (Apple Silicon and Intel), Windows, iOS, and Android.

What they don't have on their public-facing trust page yet: SOC 2 and ISO 27001 are still "in progress," they're not offline, and they don't disclose which underlying transcription model they use. Their sub-processor list points at OpenAI and Gemini, which suggests cloud Whisper plus Gemini for cleanup. If model transparency or compliance certifications matter to you, that's something to weigh.

Pick Typeless if you write a lot of long-form, you want the cleanest editing model in the category, and you don't need agent capabilities.

Viora — voice that goes beyond text

Viora is what we make. We'll be plain about what we ship today and what we don't.

The pitch is: same hold-to-talk hotkey as the others, same cleaned-up text dropping into your cursor — but the same hotkey can also trigger an agent. If your sentence is a question or a task instead of a piece of writing, Viora opens the Agent Card next to its on-screen Capsule and streams an answer. Calendar lookup via EventKit, web search, third-party connectors via Composio — all of it runs inline. "Schedule a 30-minute focus block tomorrow morning" creates the actual event. We're the only one of the four that does this today.

The other thing specific to us is the Capsule UI. A floating orb that docks to one of four corners of the screen, half-hides at the edge when you're not using it, and expands when there's something to look at. It's our visual signature.

Pricing is simple: Free with 20,000 words per month, or Pro at $20/month or $199/year. We're macOS only — Apple Silicon and Intel both supported, no Windows, no mobile.

What we honestly don't do: we're not offline (transcription is cloud-based), we don't have HIPAA compliance, we don't have a public API, and our language coverage is 28+ versus the 100+ that Wispr Flow and Typeless ship. If any of those is a hard requirement for your work, the right answer in this list isn't us.

Pick Viora if you live on macOS and you want voice to do more than dictate.

One table

How to actually choose

After all that, the decision usually comes down to four questions, in roughly this order.

Are you cross-platform? If your day moves between a Mac and a Windows machine, or you want voice on your phone too, you've narrowed the list to Wispr Flow or Typeless. Superwhisper has Windows and iOS but most of its appeal — local Whisper models — is Apple Silicon only. Viora is Mac-only.

Is offline a hard requirement? If "no audio ever leaves my Mac" is non-negotiable — for HIPAA, for legal review, or just because you want it that way — Superwhisper is the answer. The built-in feature is also offline for most languages but doesn't do post-processing.

Do you want voice to do things, or just produce text? If you keep finding yourself dictating into ChatGPT to ask it to do something with the result, voice has become a routing problem and the right answer is the agent path. That's Viora's specific bet. If voice is purely "I want to type with my mouth," any of the other three is a clean choice.

Subscription or one-time? Superwhisper at $249.99 once is the cheapest tool in the category for anyone who'll use voice for more than 16 months. Everyone else, including us, is subscription-only.

There's no universally correct answer here. We've watched friends pick all four for legitimate reasons. The mistake we see most often is people defaulting to the most-marketed option and then quietly not using it because it didn't fit their actual workflow. Spend an hour with the free tier before you commit.

FAQ

What is the best dictation app for Mac in 2026? There isn't a single answer. Built-in macOS dictation for short input. Superwhisper for offline. Wispr Flow for cross-platform and HIPAA. Typeless for long-form writing. Viora for voice that runs agent tasks. The four-question framework above is the shortest path to your specific answer.

Does macOS dictation work offline? Yes for major languages, after the language pack has downloaded once. Some less common languages still route through Apple's servers.

How accurate is Mac dictation in 2026? For clear English in a quiet room, all five options listed above are at human-level transcription accuracy. The differences show up in noisy environments, with accents, and on technical jargon. That's where personal dictionaries and stronger cleanup models matter.

Why does my Mac dictation cut off mid-sentence? The built-in feature stops listening after a few seconds of silence. There's no setting to extend the timeout. If you pause to think while dictating, dictation ends. Hold-to-talk third-party apps fix this.

Can I dictate into Cursor or VS Code? Yes. Dictation works in any text field, including code editors. Wispr Flow markets explicit Cursor and Windsurf integration; the others rely on system-level injection that works the same in any app.

Can I add custom words to Mac dictation? Not to the built-in feature — it has no public dictionary editor. All four third-party apps support a personal dictionary.

Is there a free dictation app for Mac? The built-in macOS feature is free. Wispr Flow has a 2,000 words/week free tier. Typeless has 8,000 words/week. Viora has 20,000 words/month. Superwhisper has a 15-minute trial.

A note on this guide

This page is published by the Viora team. We've tried to handle our competitors the way we'd want to be handled — name what they do better than us, link to their site, leave the choice to the reader. Pricing, platform support, and feature lists were verified against each vendor's own pages on April 28, 2026. We'll re-check when any of the products ships material changes.

If you read this far and you're on macOS and you want to try the agent angle specifically, download Viora. Free tier, no signup needed to install. If something else on this list fits your workflow better, that's the right call. We'd rather you use the tool that actually works for you than install ours and not open it.

Viora pricing · Viora vs Wispr Flow · Viora vs Superwhisper