Long replies, out loud
Draft proposals, specs and internal notes without leaving the tool you already use.
Press a hotkey anywhere and talk. The cleaned text is typed straight into whatever app is focused: email, editor, chat, terminal, browser, PDF form. No dictation window, no copy and paste.
Hi Anna,
Let's move the review to 9am. We need three things before then:
The deck
The numbers
The API review
Greeting on its own line, then a paragraph. No sign-off was added, because none was spoken.
It is not rewriting you. It punctuates, paragraphs and lists what you already said, and drops what you took back.
Long dictations transcribe while you are still talking, so only the tail is left when you stop. A ten minute dictation appears in about two seconds, and it stays about two seconds whether you spoke for thirty seconds or fifteen minutes.
Punctuation, capitalisation, filler words gone, paragraphs and lists built. Nine cleanup styles, from Literal to Email to Technical, and they work in every language rather than English only.
Only the formatting changes, never your words. In a mail client a greeting goes on its own line and the text breaks into short paragraphs; in chat it stays one casual block with no trailing period; in a terminal identifiers, paths and flags are left exactly as spoken. It will never add a greeting or a sign off you did not say.
Say "dinner at seven, actually never mind, let's do eight" and you get "Dinner at 8." You do not have to speak in finished sentences.
Names, acronyms and jargon go in a personal dictionary. Corrections you make are detected and offered back as one click suggestions. Nothing changes what you dictate next unless you say yes.
Hold to talk, or tap to talk with auto stop. Separate hotkeys for dictation, the voice instruction, voice commands, the AI agent and meetings. Your clipboard is saved and restored around every paste.
Draft proposals, specs and internal notes without leaving the tool you already use.
Dictate prose into the editor while syntax in the terminal is left exactly as spoken.
For people whose typing is painful or impossible, dictation, editing and commands are all voice.
Yes. The on-device engine supports 40 languages with real punctuation and runs on a normal processor, with no internet and no audio leaving the machine. It can be slower than the cloud mode, and that is the whole trade: you are choosing privacy over speed.
Yes. It is a desktop app, not an extension. It types into any application on your system, including native apps and terminals.
Around 100, detected automatically without you telling it which one. That is dictation coverage, and it is a different thing from the app interface, which is translated into 10 languages.
Press a hotkey anywhere and talk. The cleaned text is typed straight into whatever app is focused: email, editor, chat, terminal, browser, PDF form. No dictation window, no copy and paste.