Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

FreeFlow

FreeFlow

Free, local, private voice dictation for Windows — a full Wispr Flow replacement that costs nothing, forever. No account, no subscription, no word limits, and your voice never leaves this machine.

How it works

  1. Run FreeFlow.exe — the app window opens and a mic icon lands in the tray.
  2. Put your cursor in any text field, in any app.
  3. Hold Right Ctrl and speak — words appear while you're still talking.
  4. Release — FreeFlow re-checks the whole utterance with its accurate model and polishes the text in place (punctuation, capitalization, filler removal).
  5. Quick-tap Right Ctrl to latch hands-free; tap again (or click the pill) to stop.

While you talk, the floating pill at the bottom of the screen shows a live equalizer and the words streaming in.

Spoken commands

Say Get
"new line" / "new paragraph" line break / blank line
"scratch that" (mid-sentence) wipes what you said before it
"scratch that" (alone) deletes the last dictation

The app

  • Home — status, hotkey (click to change — press any key to rebind), lifetime stats (words, words today, speaking speed), and a test area
  • Dictation — live typing toggle, final-pass mode, tone, auto-edits
  • Dictionary — "hey gen" → "HeyGen"; teach it names and jargon
  • Shortcuts — say "my email" alone, get the full address typed
  • App Profiles — tone matching per app: casual in Discord, professional in Outlook, verbatim in terminals
  • History — every dictation, locally; Model — model manager; Settings — mic, hotkeys, autostart, AI endpoint

Settings apply instantly — no save button.

Engine architecture (all local, all CPU)

Stage Model Job
Live words Streaming Zipformer transducer (int8, 71 MB) partial transcripts ~every 100 ms while you speak
Punctuation CT-Transformer online punct (int8, 7 MB) caps + punctuation for live text
Final polish NVIDIA Parakeet TDT 0.6B v2 (int8, 652 MB) re-transcribes the utterance on release; beats Whisper large-v3 on English
Multilingual Whisper small/base (optional) 100+ languages

Command mode (select text, hold command key, say "make this more concise") works with any OpenAI-compatible endpoint — install Ollama for free local AI.

Build & verify

cd src\FreeFlow
dotnet build -c Release          # exe → bin\Release\net8.0-windows\FreeFlow.exe

FreeFlow.exe --selftest          # 33 checks: formatter, resamplers, TTS→STT, streaming partials, punctuation
FreeFlow.exe --injecttest        # live text-injection proof
FreeFlow.exe --pillpreview       # animated demo of the overlay pill
FreeFlow.exe --transcribe x.wav  # transcribe any wav

Models live in %APPDATA%\FreeFlow\models, config/history/stats in %APPDATA%\FreeFlow.

About

Free, local, private voice dictation for Windows.

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages