Granola Voice Typing is free dictation, and it's aimed at Wispr Flow

Granola spent two years doing one thing, meeting notes. In late October 2026 it shipped a second one. Voice Typing lets you hold a key, talk, and have cleaned-up text land in whatever app you're typing in. It's in beta on the Mac and Windows apps, and it's free on every Granola plan, including the $0 Basic tier.

That puts Granola straight into Wispr Flow's lane, and we live in both. We run Granola on our calls every day and wrote up a year of it in our Granola AI review. On the dictation side, we've dictated 190,767 words into Wispr Flow, and our Wispr Flow review covers how we use it. This first look is built from Granola's own Voice Typing help center, checked on November 7, 2026. We'll add accuracy notes from daily use once we've run it for a few weeks. Until then, every claim below is something Granola's docs say, and we label it that way.

What Granola Voice Typing actually does

It's push-to-talk dictation for your whole computer. Granola's docs describe it in one line, hold a shortcut, speak, and release to paste. The text goes to wherever your cursor is. Granola names Claude, ChatGPT, and Gmail as examples, and the docs say it works in any app on Mac or Windows.

• Hold to talk. The default shortcut is fn (the globe key) on a Mac, or Left Ctrl plus the Windows key on Windows. Release and it pastes.
• Hands-free mode. Double-tap the shortcut and keep talking until you press it again. Esc cancels.
• Up to 20 minutes per dictation. At 20 minutes Granola stops listening and pastes what you said so far.
• Separate from meeting notes. Granola is clear that Voice Typing doesn't start a meeting or dictate into one. It's its own feature that happens to live in the same app.
• Free on every plan. Basic, Business, and Enterprise all get it. Granola doesn't list a word cap, only the 20-minute limit per dictation.

Don't confuse it with the microphone icon on the Granola Chat bar. That one is for speaking short questions to Chat about meetings you already have, per Granola's Chat dictation page. Voice Typing is the one that types into other apps.

How to turn it on

Setup lives in Granola's desktop settings. Here's the path from the docs.

1. Open Granola on your Mac or Windows PC and click your workspace name in the bottom left to open Settings.
2. Go to Voice typing and turn on Enable Voice Typing.
3. Grant microphone access when it asks. On a Mac, also grant Accessibility, which Granola uses to catch the shortcut, read context from the app you're in, and paste. Windows skips that step.
4. Test the shortcut. Keep fn on a Mac or Left Ctrl plus the Windows key on Windows, or pick your own. Granola blocks copy, paste, undo, and app-switching combos.
5. Pick your languages under Voice typing, then Language. Until you choose, it uses up to three of your computer's languages. Auto-detect keeps each part in the language you spoke and doesn't translate.
6. Click into any text field, hold the shortcut, and talk.

One hardware tip straight from the troubleshooting page. A built-in or wired mic works best, and Bluetooth is less reliable. Dictation also needs an internet connection, so it won't work on a plane without wifi.

What the cleanup does, and what it won't touch

Raw speech-to-text is a wall of words with no commas. The reason people pay for dictation apps is the cleanup pass, and Granola's docs list exactly what its pass does.

• Removes filler words like "um" and "uh," plus accidental repeats.
• Adds punctuation and splits your thoughts into paragraphs.
• Turns spoken lists into bullets or numbers.
• Keeps your wording. Granola says it never answers or acts on what you say. It types, it doesn't chat back.

Styles. Granola's styles page sorts apps into four groups. Email (Gmail, Outlook, Apple Mail, Superhuman), work chat (Slack, Teams, Google Chat), personal chat (Messages, WhatsApp, Telegram, Signal, Discord), and everything else, which covers ChatGPT and Claude. Each group starts on Automatic and can be set to Formal or Casual, plus Excited for the work groups or Very casual for personal chat. Styles only change capitalization, punctuation, and spacing, never your words. Code editors and terminals keep their own formatting.

Dictionary. Names and jargon are where dictation falls apart, so Granola has a personal dictionary under Personalization. Each entry can be up to 50 characters and six words. It helps Granola recognize a word, but it can't swap one word for another, so it's not a text-expander. Granola also learns from spelling fixes you make right after a dictation and adds them on its own, which you can turn off. This dictionary is separate from the jargon list Granola uses for meeting transcripts.

Screen context. On by default. Granola reads text from the app you're dictating into, not screenshots, and skips password fields. It uses that to spell names right and to pick the style. Turn it off and styles turn off with it.

Privacy, what leaves your computer

Dictation hears everything you say to it, so this part matters more than the feature list. Granola's dictation history and privacy page is specific.

• It only listens while you're dictating. Holding the key or running hands-free mode is the only time the mic is on for this feature.
• What gets sent. Your audio, your dictionary, your chosen language, and, with screen context on, limited text from the app you're typing into. It goes to Granola's transcription and AI providers.
• Audio from successful dictations isn't kept. Failed or cancelled dictations keep encrypted audio for up to 24 hours so you can retry.
• Local history is your call. Store data locally (the default), delete after 24 hours, or never store locally. History only syncs if you turn on cloud storage.
• Providers don't train on it. The page says Granola's transcription and AI providers don't use your dictations to train their models.

One gap we'd flag. Granola's general model training page says Granola may use anonymized data to improve its services by default, and it doesn't mention Voice Typing either way. If that matters to you, turn off "Use my data to improve models for everyone" under Settings, then Preferences. We'd do it regardless.

Granola Voice Typing vs Wispr Flow

This is the comparison everyone searching for Granola dictation actually wants. Wispr Flow numbers come from wisprflow.ai and our own account, Granola's from its help center, both checked November 7, 2026.

Read the table two ways. On price, Granola wins outright. Free with no listed word cap beats a 2,000-word weekly allowance, and that's a real difference for anyone who dictates prompts all day. At our volume, Wispr's free tier would run out by Tuesday.

On everything else, Wispr Flow is ahead today. It's a finished product on four platforms, and we've seen its cleanup and dictionary get better the longer we use it. Granola's is a beta on two platforms. Beta means rough edges, and dictation is a feature where rough edges show up in every sentence. That's why we're not telling Wispr Flow Pro users to cancel. We're telling them to run Granola's version beside it for a week, since it costs nothing. Our Wispr Flow vs Granola comparison covers the meeting notes side of the same fight.

The catches

• It's a beta. Granola labels it that way in the docs. Expect changes, and don't build a team workflow on it yet.
• No phone app yet. If half your dictation happens on an iPhone, you'll still need something else there.
• Needs internet. Per the troubleshooting page, dictation won't run offline.
• Mac Accessibility permission. Granola needs it to paste into other apps. Some company-managed Macs block that, so check before you count on it at work.
• Bluetooth headsets are less reliable. Granola's own advice is a built-in or wired mic.
• No language count published. The docs explain language settings but don't list how many languages Voice Typing supports. Test yours before you switch.
• The dictionary doesn't do replacements. If you rely on text shortcuts, keep your text-expander app.

Who should use it

• Every current Granola user. It's already in your plan. Turning it on is free, and you'll know in a day whether you like dictating.
• People who write a lot of AI prompts. Talking a long prompt into Claude or ChatGPT is faster than typing it. Granola names both as target apps.
• Anyone who hit Wispr Flow's free cap. If 2,000 words a week isn't enough and $144 a year feels steep, Granola's free version is the obvious test.
• Windows users. It runs on both desktop platforms from day one.

And who should wait.

• Mobile-first dictators. Wispr Flow is on iPhone and Android. Granola isn't, yet.
• Teams that need admin controls for dictation. A beta isn't where you standardize a company workflow.
• Anyone who dictates offline. It needs a connection.

The bottom line

More from the same corner of the site. Our AI dictation guide covers the wider field, Wispr Flow pricing runs the paid math, Granola pricing covers the three plans, and if you're a student or a seed-stage founder, read how to get Granola free for a year before you pay for anything.