Granola takes on dictation apps with free, unlimited voice typing
The meeting notetaker now turns speech into formatted text across Mac apps, with unlimited words included even on its Basic plan.
Granola wants the writing between meetings, too. The AI meeting notetaker has launched free voice typing that puts formatted speech into the app where someone is already working, rather than asking them to compose inside a separate notes window.
TechCrunch reported the launch on October 29. Granola’s own page promises unlimited words for all users, including those on Basic. Its help center calls the feature a beta in the Mac desktop app. Windows, iOS and Android support are coming soon; the existing meeting product’s availability should not be confused with availability of this new feature.
Hold a key, speak, paste
The documented workflow starts with a text field in an email, Slack conversation, document or AI prompt. Hold the default Fn key, speak, then release it to paste the result. Double-tapping enables hands-free dictation. Each dictation can last up to 20 minutes, even though the product page advertises unlimited words overall.
Granola says it removes filler words and repeats, adds punctuation and organizes lists. It does not answer a dictated request or carry it out: speaking a prompt for ChatGPT produces text for ChatGPT. Formatting quality and recognition are product claims, not results AITrending independently reproduced.
The company’s “4x faster” presentation compares speaking at 160 words per minute with typing at 40. That is an illustrative assumption, not a measured speed improvement for every user.
Context comes with permissions
Setup requires microphone and macOS Accessibility access. Screen context is enabled by default. According to the documentation, it reads text around the cursor, selected text and nearby visible content to help with names and app-specific formatting. It can also include a window title or browser address. Granola says it reads text rather than screenshots and skips password fields.
The privacy page says audio, dictionary and language settings go to transcription and AI providers, along with limited app text when screen context is enabled. It says those providers do not train their models on dictations. Users can turn screen context off, losing app-specific styles while retaining dictation.
Local dictation history is a separate setting: by default it stays on the Mac until deleted, while encrypted audio from failed dictations is kept for up to 24 hours to support retries. Successful dictation audio is not kept, according to Granola. Local history therefore does not mean entirely on-device processing.
The immediate appeal is straightforward: existing Granola users can try cross-app dictation without another paid subscription. Whether its formatting and vocabulary handling fit a particular workflow remains something to evaluate in the beta.