You speak 3x faster than you type. SpeakFlow is a keyboard you talk to.

A macOS menu bar app that turns your voice into text — anywhere.
Press a hotkey, speak naturally, and your words appear in whatever app you're using.

Switch between modes from the Transcription tab in the settings window.

Deepgram offers a free $200 credit — no credit card required.

1. Sign up at deepgram.com/pricing
2. Create an API key in the Deepgram console
3. Paste it into SpeakFlow via the Accounts tab in settings

1. Download SpeakFlow.dmg from the Releases page
2. Open the DMG and drag SpeakFlow to Applications
3. Launch SpeakFlow — the settings window opens automatically
4. Grant Accessibility and Microphone permissions from the General tab
5. Log in to ChatGPT or add a Deepgram API key from the Accounts tab

git clone https://github.com/rezkam/SpeakFlow.git
cd SpeakFlow
swift build -c release --product SpeakFlow

# Or build the full .app bundle + DMG:
bash scripts/build-release.sh 0.1.0

Requires macOS 15+ and Swift 6.1+ (Xcode 16.4 or later).

SpeakFlow needs two permissions, both granted from the General tab in settings:

• Accessibility — required to type transcribed text into any app. Clicking "Grant Access" opens System Settings where you toggle SpeakFlow on.
• Microphone — required to hear your voice. Clicking "Grant Access" shows the macOS permission dialog.

Permissions are never requested automatically on launch — you choose when to grant them.

SpeakFlow is self-signed (ad-hoc), so macOS ties the accessibility trust to the exact binary. If you rebuild the app or install a new version and accessibility stops working:

1. Open System Settings → Privacy & Security → Accessibility
2. Remove the old SpeakFlow entry (select it and click the minus button)
3. Re-add SpeakFlow.app from /Applications (click the plus button)
4. Restart SpeakFlow

This is required because macOS revokes accessibility trust when the binary changes for apps without an Apple Developer ID signature.

• Real-time streaming transcription — words appear as you speak with Deepgram Nova-3; interim results refine in-place using smart diff (only changed characters are retyped, no flickering)
• Batch transcription — record first, transcribe after with GPT-4o via ChatGPT
• On-device voice activity detection — a neural network model runs locally on Apple Silicon to distinguish speech from silence in real-time; no audio leaves your machine until speech is confirmed
• Automatic turn detection — when you stop speaking, silence is detected and the session ends automatically (works in both modes — local VAD for batch, server-side for streaming)
• Smart chunking — in batch mode, audio is split at natural sentence boundaries detected by silence analysis
• Noise filtering — silent and noise-only chunks are filtered before transcription, saving API calls and improving accuracy
• Universal text insertion — transcribed text is typed into the focused app via macOS Accessibility
• Launch at login — runs quietly in the menu bar

1. Capture — 16 kHz, mono, 16-bit PCM from the system microphone
2. Stream — raw audio is sent over WebSocket to Deepgram Nova-3 (English, monolingual)
3. Interim results — partial transcriptions appear immediately as you speak
4. Smart diff — when text updates, only the changed suffix is retyped (common prefix is preserved)
5. Final results — server finalizes each utterance with punctuation and formatting
6. Auto-end — after 5+ seconds of server-detected silence following speech, the session ends

1. Capture — 16 kHz, mono, 16-bit PCM from the system microphone
2. Voice activity detection — a lightweight neural model runs on Apple Neural Engine, classifying 32ms frames as speech or silence
3. Speech segmentation — hysteresis thresholds with 3-second silence debounce avoid false ends during natural pauses
4. Chunking — audio is buffered and sent at the configured interval, waiting for natural pauses
5. Auto-end — after 5+ seconds of confirmed silence following speech, the session ends
6. Transcription — speech chunks are sent to the Whisper API; results arrive in order and are typed into the active app

Apache License, Version 2.0