Stop switching into"typing mode."
Kufuri Typing is not just dictation.
Speak roughly, let AI clean it up, and insert it when you are ready.
Rewrite selected text or write with context from the page, screen, or file you are already looking at.
The app itself is free · Start without an API key if you are already signed in to ChatGPT · macOS 14+ · Apple Silicon / Intel
Nothing is inserted the moment you speakRewrite selected textUse pages, screens, and PDFs as contextThe problem with dictation is not speed.
It is having to switch into input mode.
The painful part is not typing speed. It is the mental switch you have to make before you start writing. Kufuri Typing separates speaking from inserting text, so that switch happens less often.
Typical dictation
It sounds convenient, but it still keeps you inside an "input mode."
You have to speak in finished sentences- Your words are inserted immediately, which makes rough thinking awkward
- It is weak at rewriting existing text or using on-screen context
- You still have to brace yourself and think, "okay, time to type"
Kufuri Typing
It changes the flow so you can use it in the middle of thinking, not only when you are ready to type.
- You can speak roughly
- AI can clean it up, but nothing is inserted until you confirm
- It can work from existing text, pages, screens, and files
- So you can write and revise with your voice without breaking flow
Speak roughly
You do not need polished sentences. Fragments, filler words, and half-formed thoughts are all fine.
Let AI clean it up
Tone instructions like "make it polite" or "turn this into bullets" can be part of the same spoken draft.
Insert it when you want
Because nothing is inserted until you confirm, you can use it without breaking your train of thought.
What makes it different comes down to
these three things.
It does more than reduce the friction before you start writing. Kufuri Typing treats drafting, rewriting, and context-aware writing as one continuous workflow.
Speak roughly, insert later
You do not need to dictate polished sentences. The core of Kufuri is that speaking and inserting text are separate steps, so the mental friction drops before you even start.
- Start recording with ⌘J
- Send rough spoken language to AI as-is
- Confirm insertion with ⌘I
Rewrite what is already selected
It is not only for writing from scratch. You can take selected text and make it more polite, shorter, or more structured without leaving the current app.
- Works directly on selected text
- Easy to shift between email, chat, and notes
- Say things like "make it shorter" or "make it more polite"
Use what you are already looking at
Pages, screenshots, files, and PDFs can all become context. That is the decisive difference from ordinary dictation.
- Use page context from Chrome, Safari, Brave, and Edge
- Respond based on screenshots and reference materials
- More than transcription: it rewrites with understanding
It reduces not only the friction before you start writing,
but also the places where you get stuck mid-write.
Kufuri Typing does not stop at the first draft flow. It is designed to stay useful in day-to-day work, with the supporting features that make that possible.
Floating bar
Start from a minimal UI at the top or bottom of the screen, no matter which app you are using. You do not have to switch apps just to enter text.
Rewrite selected text
It is not only for creating new text. You can also revise what already exists with your voice. Make emails more polite, turn notes into bullets, or shorten status updates on the spot.
Use pages, screens, and files as input
Pass screenshots, browser pages, files, and PDFs to AI. You can write from what is already on screen, which makes replies and summaries much faster.
Custom instructions by app or site
Set output style by context, such as "keep it concise in this app" or "use a review tone on this site." Your writing does not have to sound the same everywhere.
History timeline
Look back at audio, rewritten text, screenshots, and context in a timeline. It is designed not just as a one-shot input tool, but as a working log.
Choose your speech engine
Switch between standard, high-accuracy, and Voxtral Realtime. Choose the right balance of latency, accuracy, and on-device behavior for your environment and language.
5 AI providers
Choose from ChatGPT / OpenAI / Anthropic / Google / xAI.
4 supported browsers
Use page context from Chrome / Safari / Brave / Edge.
57 speech locales
Support extends far beyond Japanese, across many recognition locales.
14 UI languages
Choose the app UI language and prompts from 14 languages.
Apple Silicon / Intel
Supports both on macOS 14+, with architecture-specific downloads available.
It does not record on its own.
It does not capture on its own.
The biggest concern tends to be "this looks useful, but is it watching all the time?" Kufuri Typing makes those boundaries explicit from the start.
Recording happens only when you press ⌘J
It records only when you explicitly start it. It is not designed around background recording.
Screen capture happens only at ⌘⇧I
Screenshots and page context are sent only when you deliberately invoke them. There is no always-on screen monitoring.
Saved data stays on your Mac
Text, history, screenshots, and audio are stored locally. You can change retention settings and disable audio saving.
Keychain-managed + signed and notarized
API keys are stored in macOS Keychain. The app is signed and notarized with Apple Developer ID.
Note: With the standard speech engine, audio may be processed by Apple servers. With the high-accuracy engine, recognition runs on-device. During AI formatting, the required text or images are sent to the provider you selected.
Answers to the questions people usually ask
before trying it.
Pricing, permissions, storage, and the other things people want to know first.
The app itself is free. If you are already signed in to ChatGPT, you can get started without an API key. If you use OpenAI / Anthropic / Google / xAI directly, their own usage terms apply.
Accessibility permission is required. It is used for text insertion and global shortcuts. Screen Recording is required when you use screenshots or page context with AI formatting. macOS asks for microphone permission when you record.
Text, history, screenshots, and audio data are stored on your Mac. You can change retention settings and disable audio saving. Only the necessary content is sent to the selected provider when AI formatting runs.
No. Recording starts and stops with ⌘J, and screen capture is used only at the moment you invoke ⌘⇧I. It is not designed to keep recording or watching your screen in the background.
That depends on the speech engine you choose. The standard engine may rely on Apple servers. The high-accuracy engine runs on-device. AI formatting itself uses the AI provider you selected.
Yes. It supports both Apple Silicon and Intel on macOS 14+. Chrome and Edge can auto-detect architecture. If you are on an Intel Mac, use the Intel download.
Yes. The UI supports 14 languages, and speech recognition supports 57 locales. You can switch both interface language and recognition language in settings.
Before you reach for the keyboard,
try drafting with your voice first.
Cutting just the "let me get into writing mode" moment makes input feel much lighter. The easiest way to feel it is to try one email or chat reply.
The app itself is free · Start without an API key if you are already signed in to ChatGPT · Accessibility permission is required first