100% offline option
Local whisper.cpp transcribes on your machine. The speech engine and model download automatically on first use (one-time, needs internet) — after that it works with the Wi-Fi off.
Type with your voice.
Anywhere.
Hold a key, speak, and clean text appears in any Windows app — free forever, open source.
Free forever · No account · No word caps · Open source (MIT)
v0.1.0 · Windows 10/11 (x64) · ~105 MB · unsigned installer — SmartScreen may warn (More info → Run anyway)
01 — How it works
Three moves. No window switching.
-
Hold Ctrl+Alt
The Flow Bar appears. The hotkey is rebindable, and there’s a hands-free toggle mode too.
-
Speak naturally
Ums, pauses, run-on thoughts — say it the way you’d say it out loud.
-
Release
Clean, punctuated text lands in whatever app has focus — editor, email, chat, terminal.
02 — Features
Everything you need. Nothing metered.
Bring your own keys
Prefer cloud speed? Plug in your own Groq or OpenAI Whisper key. Your keys, your data — stored encrypted with Windows DPAPI.
AI cleanup — even fully local
Optional cleanup removes filler words and fixes punctuation and formatting, via OpenAI, Groq, or a fully local Ollama model.
Voice-trigger snippets
Say “my email” and the full expansion is typed instead. Define any trigger → text pair you like.
Scratchpad
A global shortcut opens a quick notes pad — capture a thought by voice without leaving what you’re doing.
Insights
Words per minute, streaks, and per-app stats — see where and how much you actually dictate.
History, clipboard-safe
Every dictation is kept in a local history. Your clipboard is saved before each paste and restored right after.
Push-to-talk or hands-free
Hold to talk for quick bursts, or flip into toggle mode for longer dictation sessions.
03 — Privacy
Your voice stays yours.
- In local mode, audio never leaves your PC — speech-to-text runs on your machine via whisper.cpp.
- No account, no sign-up, no telemetry. Nothing is sent anywhere unless you add your own API key and choose a cloud provider.
- API keys are stored encrypted with Windows DPAPI and never shown in plain text.
- It’s all MIT-licensed and open. Every claim about CodeFlow on this page is checkable in the code.
Don’t trust us. Read the source.
04 — Compare
How CodeFlow stacks up
| Feature | CodeFlow | Wispr Flow | Superwhisper | Dragon Professional | Windows Win+H |
|---|---|---|---|---|---|
| Price | Free forever | $12/mo billed annually ($15 monthly) | From $8.49/mo | ~$699 one-time | Free, built in |
| Free-tier word caps | None | 2,000 words/week | None (small local models; 15-min Pro trial) | No free tier | None |
| Works offline | Yes — 100% local option | No — cloud speech-to-text | Yes — local models | Yes | No — requires internet |
| Account required | No | Yes | For paid features | License activation | No |
| Open source | Yes — MIT | No | No | No | No |
| AI cleanup | Yes — cloud or fully local (Ollama) | Yes — cloud | Yes — via modes | No | No |
| Windows-native | Yes — built for Windows | Yes (also Mac & mobile) | Windows version available (Mac-first) | Yes — Windows only | Yes — built in |
Pricing and limits as of July 2026, from each product’s official pricing page. Dragon Professional pricing via major retailers (Nuance lists pricing through sales). All product names belong to their respective owners.
05 — Why free?
Why is it free?
CodeFlow is built and maintained by one developer — contributions welcome. It’s MIT-licensed because we think dictation shouldn’t rent your own voice back to you — speaking is not a metered resource, and a tool this personal shouldn’t come with an account wall.
Running costs are small and covered out of pocket today — donations, once they’re set up, will keep it that way. There’s no venture funding to pay back, so there’s no pressure to add a paywall later.
In the future there may be an optional hosted plan — priced around covering API and infrastructure costs — for people who’d rather not manage their own API keys. But the app itself stays what it is today: free, uncapped, account-free, forever.
06 — Support the project
Keep it free for everyone
If CodeFlow saves you a subscription, consider tossing a little of that back.
Donation links are being set up — for now, starring the repo and filing issues is the best support.
Not into donating? Starring the repo, sharing it with a friend, or filing a good bug report helps just as much.
07 — FAQ
Questions, answered plainly
Is it really free?
Yes. Free forever, MIT-licensed, no word caps, no account, no trial that expires, no Pro tier to upsell you into. If a hosted convenience plan ever exists, it will be optional — the app stays free.
Do I need an API key?
No. Local mode works out of the box — the speech model downloads automatically on first use and runs entirely on your PC. If you’d rather use the cloud, a Groq key is a fast, currently-free cloud option; OpenAI Whisper works too.
Is my voice uploaded anywhere?
In local mode, never — audio is processed on your machine and nothing is sent anywhere. There is no telemetry. If you add your own API key and choose a cloud provider, your audio goes directly from your PC to that provider (Groq or OpenAI) under your key — CodeFlow has no servers of its own.
Why does Windows warn during install?
The installer isn’t code-signed yet (signing certificates cost real money), so SmartScreen shows a warning on first run. Click “More info” → “Run anyway”. The whole app is open source — you can read the code or build the installer from source yourself if you prefer.
Is there a Mac or Linux version?
No — CodeFlow is Windows-first (Windows 10/11, x64), and nothing else is planned yet. We’d rather be excellent on one platform than mediocre on three.
How is this different from Win+H?
Windows’ built-in voice typing sends your speech to Microsoft’s online services and stops working without internet. CodeFlow can run 100% offline, cleans up filler and punctuation with optional AI, and adds snippets, a scratchpad, history, and stats — all behind one always-ready hotkey. One honest limit: apps running as administrator can’t receive simulated paste from a normal-privilege app — run CodeFlow as administrator if you need to dictate into elevated windows.
How do I get help?
Open an issue on GitHub. Bug reports, feature requests, and questions are all welcome there.