Windows · 100% offline

Dictate anywhere.
Speech never leaves this PC.

Hold Ctrl+B, speak, and cleaned text pastes into the app you are already using. Recognition runs with whisper.cpp on-device. No account. No telemetry. No cloud API.

Current-user installer · no admin · unsigned until we add a certificate (SmartScreen may warn)

See it work

Push to talk

Hold Ctrl+B, speak, release. Text lands in the focused window.

Long listen

Press Ctrl+B twice for longer takes. Space or Esc to stop.

On-device Whisper

base.en and Silero VAD ship in the installer. No download at first run.

Rules before AI

Fillers, times, and vocabulary are cleaned locally even without Qwen.

How it works

  1. Install the NSIS .exe for your Windows user.
  2. Click a text field in any app.
  3. Hold Ctrl+B and speak.
  4. Release. Cleaned text is pasted. The mic is not left open.

Privacy, by default

Audio, transcripts, and clipboard contents are never uploaded. History and recordings stay off unless you turn them on. There is no Ollama, no OpenAI, and no Python runtime in the shipped app.

Optional local Qwen

Want extra punctuation cleanup? Download Qwen3-0.6B-Q8_0.gguf in a browser, copy it to %LOCALAPPDATA%\LocalFlow\models, then restart. LocalFlow never fetches models itself.

Full Qwen setup