Voice to text that knows your app.
Open-source macOS dictation powered by whisper.cpp. Audio never leaves your Mac.
No subscription, no lock-in.
git commit "fix perfectly align pulsing mic animation"
Your terminal and your editor need different things
Most dictation tools paste the same way everywhere — breaking terminals and slowing down editors. Steno detects your active app and picks the right method.
Terminals: clipboard paste
Instant insertion, safe for zsh/bash
Editors: direct typing
Natural typing, triggers autocomplete
See it in action
Built for developers who talk to their Mac
How it works
App-Aware Insertion
Detects your active app and picks the safest insertion method. Terminals get paste. Editors get direct typing.
Local Transcription
Powered by whisper.cpp. High-accuracy transcription runs entirely on your hardware.
Global Hotkeys
Option hold-to-talk for quick dictation. Configurable function key for hands-free toggle.
Your data, your rules
Private by Default
whisper.cpp transcribes on your hardware. No server, no upload. The code is open — verify it yourself.
Free Forever
Open source under MIT. No subscription, no lock-in. Community-driven and contributor-friendly.
Local Text Cleanup
Strips filler words, whisper artifacts, and false starts on-device. Style profiles let you set tone per app.
Quality of life
Transcript History
30-day searchable history of everything you've dictated.
Personal Lexicon
Custom vocabulary, style profiles, and snippet expansion. Steno remembers your words.
Auto-Pause Media
Automatically pauses your music, podcasts, or videos while you dictate, then picks back up where you left off.
vs. the alternatives
Why you'll ditch your current dictation tool
Most voice-to-text tools are subscription-based, cloud-dependent, and closed source.
Subscription tools
Wispr Flow charges $15/mo. superwhisper Pro is $8/mo. That's $100-180/year for voice typing.
Free & Open Source
Steno is free and open source. $0/yr. Forever. Transcription and text cleanup both run locally — no API key needed.
Cloud-based voice tools
Wispr Flow and others send audio to remote servers. Some offer a 'privacy mode' — as a paid upgrade.
Private by Design
Transcription runs on your Mac, period. Open source means you can read every line to confirm.
One-size-fits-all paste
Most tools paste the same way in every app. This breaks terminals and slows down editors.
App-Aware Insertion
Steno detects your target app and picks the right insertion method. The only dictation tool that does this.
Up and running in 5 minutes
Five terminal commands. One Xcode build.
Clone the repo
Grab the source code directly from GitHub.
git clone https://github.com/Ankit-Cherian/steno.git && cd stenoBuild whisper.cpp
Build the whisper.cpp transcription engine from source.
git clone https://github.com/ggerganov/whisper.cpp vendor/whisper.cpp && cd vendor/whisper.cpp && git checkout v1.8.3 && cmake -B build && cmake --build build --config Release && cd ../..Download a model
Download the small English model for whisper.cpp.
cd vendor/whisper.cpp && ./models/download-ggml-model.sh small.en && cd ../..Download the VAD model
Voice Activity Detection filters silence and background noise. Enabled by default.
cd vendor/whisper.cpp && ./models/download-vad-model.sh silero-v6.2.0 && cd ../..Build and run
Generate the Xcode project and run the build. Grant permissions when prompted.
xcodegen generate
open Steno.xcodeprojStart dictating
Hold ⌥ Option and speak. Release to transcribe. That's it.
Why I built this
Every macOS dictation tool I tried had the same problems: a monthly subscription, audio uploaded to the cloud, and no way to verify what's actually happening with your voice data.
I built Steno to fix that. Transcription runs locally through whisper.cpp, it costs nothing, and the code is right there on GitHub. Open source under MIT.
Ankit Cherian
Creator of Steno