Granola now offers free voice typing in beta across Mac apps, but there's a setting worth checking before you dictate your first work email. Screen context is on by default, letting the tool read text from the app you're writing in to help with spelling and formatting.
That could make dictation more useful when a message includes unfamiliar names. It also means your spoken words aren't necessarily the only information being processed. The practical choice is whether you want that extra context, not simply whether you'd rather talk than type.
A separate tool for writing
The voice-typing product page promises unlimited words for all Granola users, including the free Basic plan. It's available on Mac, with Windows, iOS and Android support still described as coming soon.
According to the setup guide, the feature is in beta. Enable it in Settings > Voice typing, then grant microphone and Accessibility permissions. Hold the default Fn shortcut, speak, and release it to paste the result where your cursor is. You can change that shortcut.
This isn't a way to start a meeting recording. It's separate from meeting transcription and cleans up your own speech, removing filler and adding punctuation or structure. It doesn't carry out spoken requests: dictating a prompt into an AI app enters the words rather than executing the request itself.
Decide how much context to share
Granola's spelling and personalization guide explains that screen context can help recognize a recipient's name or adapt formatting to an email, chat or one-line field. Its dictation dictionary is separate from the jargon list used for meetings, so don't assume an existing custom list carries over.
The dictation privacy documentation makes an important distinction: history stays on your Mac, but processing isn't entirely local. Audio, your dictionary and chosen language go to transcription and AI providers. With screen context enabled, limited app text goes too, potentially including nearby messages, selected text and the browser address.
Granola says it reads through macOS Accessibility, doesn't take screenshots and skips password fields. You can switch off Screen context in Settings > Voice typing. Dictation still works, although styles and app-specific formatting turn off. Its providers don't use dictations to train their models.
For a first try, use a low-stakes draft and compare the result with what you meant, especially names, numbers and commitments. If you don't want surrounding conversations used as dictation context, switch Screen context off before speaking. Free dictation is useful only if both its output and its settings fit the work you're doing.