Welcome to OpenFlow — the open-source, local-first voice dictation app. This guide walks you from install to your first dictated sentence in a few minutes.

1. Download and install

Grab the latest build for your operating system from the Releases page.

  • Windows: Run the OpenFlow Setup .exe installer. OpenFlow launches automatically when setup finishes.
  • macOS: Open the .dmg, drag OpenFlow to your Applications folder, and launch it. On first run you may need to allow the app in System Settings > Privacy & Security.
  • Linux: Use the .AppImage (make it executable with chmod +x and run) or install the .deb package.

No account, no API key, and no internet connection are required.

2. Grant microphone access

On first launch, OpenFlow asks for microphone permission. Allow it so the app can capture your speech.

  • Windows: Settings > Privacy & Security > Microphone — make sure OpenFlow is allowed.
  • macOS: System Settings > Privacy & Security > Microphone — enable OpenFlow.
  • Linux: Depends on your desktop environment and PulseAudio/PipeWire setup.

3. Pick a transcription model

Open Settings > Models and configure a local transcription engine.

  • Windows users: Click Install next to Parakeet. OpenFlow downloads the pinned sherpa-onnx runtime and multilingual INT8 model (about 510 MB) and keeps a warm local server ready between dictations. Parakeet gives near-instant transcription.
  • All platforms: Use the built-in Whisper engine. The Tiny model (75 MB) gives the lowest latency. Download it from Settings > Models, then click Select to activate it.

For full model details, see the Model Setup guide.

4. Set your hotkey

Open Settings > General and choose a global shortcut. The default is Ctrl+Win for push-to-talk. Hold the key while you speak, then release it to transcribe, clean, and paste.

Pick a combo that does not collide with other apps. If a hotkey stops working, another app may have grabbed it — switch to a different one.

5. Dictate anywhere

  1. Focus the app where you want text to appear (an editor, email, chat, browser field — anything).
  2. Hold your hotkey and speak naturally.
  3. Release the hotkey. OpenFlow transcribes your speech, cleans it up, and pastes the result into the focused app.

That's it — you're dictating. Read on to make OpenFlow truly yours.

6. (Optional) Enable LLM cleanup

For smarter text refinement, run a local llama.cpp server and point OpenFlow at it:

llama-server -m path/to/gemma-4-e2b-it-Q4_K_M.gguf --port 8080

Then in Settings > Models, set the cleanup provider to Local llama.cpp server and the base URL to http://127.0.0.1:8080/v1. If the server is unavailable, OpenFlow automatically falls back to rule-based cleanup.

7. Make it yours

  • Personal Dictionary: Settings > Dictionary — add names, jargon, and acronyms so they transcribe correctly.
  • Snippets: Settings > Snippets — create shortcuts that expand into full text blocks as you dictate.
  • Style Transforms: Rewrite dictated text as Professional, Casual, Concise, or Bullet points.
  • Scratchpad: Quick persistent notes without leaving the flow.

Where to go next

  • Model Setup — deep dive on Parakeet, Whisper, and the Gemma cleanup model.
  • Troubleshooting — fixes for common issues.
  • Architecture — how the transcription and cleanup pipelines work under the hood.
  • Privacy — our local-first, zero-telemetry commitment.