For developers · Free · On-device

Never touch a keyboard again.

DijiFlow Dictate turns your voice into text in any app, so you prompt Claude, Cursor, VS Code and the terminal about 4× faster, drive several tools at once, and write better prompts. 100% on your machine. Free forever.

🎤 Start free, no card, no email Download free

Free forever · No account · 90+ languages · 100% on-device · macOS + Windows

Download + pricing

Start free. Go Pro when you're ready.

Free forever, no account. Try every Pro feature for 30 days, no card, no email, then pick the term that fits.

Free

£0.00 forever

No account, no card. Yours to keep.

Download for MacmacOS 14+ · Apple Silicon Download for WindowsWindows 10 / 11

Includes a 30-day free trial of Pro — every feature unlocked, no credit card.

Unlimited dictation. No credits, no caps.

Usage limitsUnlimited dictationUnlimited · no credits, no capsCredits or per-minute caps
Hotkey slots1 slot3 slotsUsually web-based — no native hotkeys
File & video length3 minUnlimitedMixed — often capped
File & video transcriptionUsually included
Speaker labels & timestampsUsually an added cost
Vocabulary support · 29 languagesVery limited or none
GPU / Neural Engine + largest modelsRuns in the cloud — data leaves your device
All 90+ languagesOften limited
100% on-device, no telemetryAudio processed in the cloud
Works offlineInternet required

Why not just the built-in mic?

You're already using voice? You're still missing most of it.

Every editor and OS has some mic button now, so why DijiFlow? Because a built-in dictation toggle and a real, on-device voice layer aren't the same thing. If you're already talking to your AI, good instinct. Here's everything you're leaving on the table.

Built-in mic / OS dictation
DijiFlow Dictate
Speed
General model, often a cloud round-trip and a noticeable lag
Tuned to your GPU and Neural Engine, keeps up with you, about 4× faster than typing
Accuracy on code
Fumbles function names, APIs, acronyms and CamelCase
Best-in-class models + your own vocabulary for jargon, libraries and names
Languages
A handful, tied to the OS, usually English-first
90+ languages, every one of them 100% on-device
Privacy
Audio is usually sent to Apple / Microsoft / Google or the app's servers
100% on your machine. Your voice and code never leave the device, works offline
Where it works
Only in that one app, and only if it has a mic at all
Every app: terminal, editor, browser, chat, email, docs, anything with a cursor
How you trigger it
Hunt for a mic icon, click it, click again to stop
One global hotkey, hands never leave the keyboard, no clicking
Several at once
One app at a time
Drive multiple terminals in parallel, talk to the next while the others think
Make it yours
Take it or leave it
Your hotkey, models, push-to-talk or toggle, custom vocabulary
Already talk to your AI? DijiFlow just makes it 4× faster, private, and everywhere.
Same habit you already have, minus the lag, the clicking, and the cloud.
Try it free for 30 days →

The unlock

Speak faster. Say more. Ship better.

Voice isn't just about 4× faster than typing. It makes your prompts better. You give the AI the full picture: the constraints, the edge cases, the "oh, and also" you would never bother to type. Richer prompts in, better code out, with fewer back-and-forths. And you do it hands-free, eyes on your screen, coffee in hand.

⚡~4×Faster in

You speak ~150 words/min. You type ~40. The slowest part of prompting just disappeared.

🗣️You say more

Talking is effortless, so you actually explain the context, constraints and intent, the prompt you would never type out.

✅Better output

Fuller prompts mean more accurate edits, fewer wrong guesses, and far less re-prompting. Quality compounds.

🐞
Describe the bug as you read it

Narrate the stack trace while your eyes stay on it. Nothing to memorize.

🎨
Fix the UI while you watch it

Look at the page and just say what's off: "tighten the spacing, make it sticky."

🔀
Steer a diff hands-free

Approve, reject or redirect a change by voice as you scroll it.

🧠
Think out loud

Ramble the whole idea; send one clear, complete prompt instead of a trimmed one.

↩️
Answer mid-task

The agent asks something, reply in two seconds and keep your flow.

📄
Anywhere text goes

Same hotkey in the terminal, editor, browser, chat, email and docs.

The math

It pays for itself before lunch.

Developers now spend 3 to 4 hours a day prompting AI. Cut the slowest part, the typing, and you get most of that back.

~4×
faster input. Speak ~150 wpm vs. type ~40
2 to 3 hrs
handed back every day, 10 to 15 hrs a week
Day 1
the time saved out-earns the Lifetime price the first day
3 to 4 hrs/day prompting→typing cut ~4×→2 to 3 hrs/day back→10 to 15 hrs/week

At any developer's hourly rate, a dozen hours a week is worth many times a one-time $39 Lifetime, so it clears its own cost the very first day.

Free for 30 days. No card. No email.Try DijiFlow free →

Works with everything

One hotkey. Every window.

DijiFlow types into the active app, so it works in every AI coding tool, and in every other app you live in. Nothing to integrate.

VS CodeWhatsAppClaude CodeGmailCursorMicrosoft WordTerminalSlackChatGPTExcelCodexOutlookNotionWindsurfDiscordGeminiReplitTelegramJetBrainsGoogle DocsZedTeamsWarpHotmailAntigravityLinearNeovimPerplexityXcodeObsidianVimJiraIntelliJMessagesPostmanTrelloAndroid StudioPowerPointGitHubSignalSublime TextConfluenceCopilotZoomPyCharmApple NotesDockerRedditAiderLinkedIniTermBearSafariFigmaVS CodeWhatsAppClaude CodeGmailCursorMicrosoft WordTerminalSlackChatGPTExcelCodexOutlookNotionWindsurfDiscordGeminiReplitTelegramJetBrainsGoogle DocsZedTeamsWarpHotmailAntigravityLinearNeovimPerplexityXcodeObsidianVimJiraIntelliJMessagesPostmanTrelloAndroid StudioPowerPointGitHubSignalSublime TextConfluenceCopilotZoomPyCharmApple NotesDockerRedditAiderLinkedIniTermBearSafariFigma

How your day changes

Run the whole desk hands off the keyboard.

🖥️
Drive several terminals at once

Open a row of terminals, speak to one, hop to the next. While they think, you're already dictating the next prompt. No idle hands, no waiting.

⌨️
Put the keyboard down

Lean back, coffee in hand, and build a whole feature by talking. Fewer keystrokes is easier on your wrists and frankly more pleasant. People who switch don't go back.

🔒
100% on your machine

Transcription runs on your device, on the Apple Neural Engine or your local GPU. Your voice and your code never touch a server. Works fully offline.

🏢
Safe at work

Because nothing leaves the device, it doesn't conflict with company privacy or data policies. Fine on confidential code, even in the enterprise.

🆓
Free forever

The Free tier is genuinely free, no card, no account. Pro just unlocks the bigger models, longer files, speaker labels and more.

🖥️
Mac & Windows

Native on both, with the same voice hotkey everywhere. One Pro licence activates on both.

Zero-risk trial

Try every Pro feature free for 30 days. No card. No email.

Unlock everything for 30 days with nothing to sign up for. Love it? Continue with Pro. Don't? It reverts to Free on its own, nothing to cancel, nothing charged.

✓ No credit card ✓ No email or account ✓ Full Pro for 30 days ✓ Reverts to Free, no cancelling ✓ 100% private
Get DijiFlow free →
Get the app

Download and start talking

Lightweight app. Speech models download on first use, then fully offline. Free, no account.

No credits · No limits

Download for Mac

v7.17

macOS 14 Sonoma+ · Apple Silicon

  • Unlimited on-device dictation
  • All 90+ languages, offline
  • No account, no card, no telemetry
  • 30-day Pro trial, then Free, nothing to cancel
Download for MacmacOS 14+ · Apple Silicon

Windows

Windows 10 / 11v7.16

Native build for Windows 10 & 11 — runs on AMD, Intel and NVIDIA GPUs. NVIDIA needs the latest GPU driver + CUDA runtime; AMD & Intel just need up-to-date drivers.

  • Unlimited on-device dictation
  • All 90+ languages, offline
  • No account, no card, no telemetry
  • 30-day Pro trial, then Free, nothing to cancel
Download for WindowsWindows 10 / 11
View the changelog
macOS 14+ · Apple Silicon Windows 10 / 11 · AMD · Intel · NVIDIA NVIDIA: latest driver + CUDA Mic + Accessibility 300 MB–6 GB model, downloaded once