Voice typing in Gmail, Slack, Docs, Word, Notion and Obsidian

No. Google documents its voice typing for Google Docs and Slides speaker notes, not for the Gmail compose window. What does type into Gmail is the dictation built into Windows and macOS, or a desktop client such as VibeVoice: click into the reply, press the hotkey, speak, and punctuated text lands in the message. Nothing is installed into Gmail, nothing gets access to your mailbox, and you press send yourself.

Outlook and Thunderbird work the same way, because no mail client is treated specially.

Not as typed text. Slack’s clips post a recording of up to five minutes, and a transcript can be generated for the clip, but what arrives in the channel is still audio. VibeVoice types into the composer instead, so you send an ordinary message people can skim, search and quote. It works in channels, threads, DMs and message edits, in the desktop app and the browser, with no Slack app to approve, and it never posts for you: you press enter.

At the edge of Google’s own documents. Google lists it for the latest Chrome, Edge and Safari, in Docs and in Slides speaker notes, and you say punctuation out loud: “comma”, “period”, “new paragraph”. Its voice commands work only in English. VibeVoice types into any browser, Firefox included, and into every other window, with punctuation taken from the way you speak. What Google’s tool does that this one does not is edit by voice: “select the last sentence”, “bold that”.

If all your writing happens in Word, use Dictate: it comes with a Microsoft 365 subscription and takes formatting commands like “bold” and “new line”, which VibeVoice does not. The case for a separate client is everything outside Word. The same hotkey types into Outlook, a browser form and a chat window, and it needs no 365 subscription, so it works with a perpetual licence, an older Word, or LibreOffice. We have not benchmarked Dictate and make no claim about which transcribes better.

None is needed. VibeVoice types into the block your cursor is in, the way a keyboard does: a text block, a toggle, a table cell, a page title, a comment, in the desktop app or a browser tab. Notion sees keyboard input, so there is no workspace permission and nothing for an admin to approve. The flip side is that it cannot create pages or fill a database, and headings and bullets stay Notion’s slash commands: make the block, then dictate into it.

No. Obsidian’s own Audio recorder saves a recording into the note, not text. VibeVoice runs outside Obsidian and types punctuated prose into the open note, so nothing is written into .obsidian, and Sync or a git vault picks the text up as if you had typed it. Your notes stay on your machine; the client cannot read them. The audio you dictate does leave it, because it is transcribed on our servers. If that rules it out, you want a tool that runs the recognition locally.

Any window with a text cursor: Outlook, Teams, Jira, a browser form, a terminal, a PDF form. There is no list of supported apps, because the client types into whatever has focus rather than integrating with each one. The exceptions are programs that deliberately reject synthetic input — some games, some remote-desktop and virtual-desktop sessions, and some hardened enterprise clients.