Press a shortcut, speak, and your words appear as text. Choose from multiple transcription engines, add AI post-processing, and everything runs locally on your Mac. One-time purchase, no subscription.
Switch between Apple Speech, OpenAI Whisper, and NVIDIA Parakeet ONNX in one click.
Each engine has different strengths: Apple Speech for zero-setup, Whisper for multilingual accuracy, Parakeet for blazing-fast English.
Go beyond raw transcription. Each mode runs your text through an LLM with a custom prompt.
Connect any OpenAI-compatible provider, Anthropic, or use Claude CLI with your existing subscription.
Assign a dedicated global shortcut to each mode. One keystroke to switch and start recording.
Configure shortcuts in Settings. Supports any key combo with ⌘, ⌃, ⌥, ⇧ modifiers.
Transcribed text is placed on your clipboard and pasted into the active app automatically.
Works with any app that accepts ⌘V. Toggle auto-paste on or off in settings.
Audio never leaves your machine. All transcription happens entirely on-device.
No cloud APIs, no data collection, no accounts. Your voice data stays on your Mac.
The audio engine stays pre-warmed in the background. Recording starts the instant you press the key.
No spin-up delay, no loading screens. Press the shortcut and speak immediately.
Distinct sounds for start, stop, and transcription complete so you always know what's happening.
Pick from 14 macOS system sounds per action, or turn them off. You hear exactly when recording starts and when your text is ready.
Connect OpenAI, Anthropic, Claude CLI, or any compatible API to power your modes.
Use your existing subscriptions. Each mode can use a different provider and model.
Appearance theme, overlay toggle, auto-paste, Python venv path — everything is tuneable.
A native macOS settings window with live preview. No config files to edit.
From any app, at any time.
From any app — a text editor, browser, Slack, terminal. The menu bar lights up and a floating overlay appears.
Talk at your normal pace. The overlay shows the live audio level. Take as long as you need — there's no time limit.
Hey, can you review the pull request I just opened? I've refactored the auth middleware to fix the session bug.
Press the shortcut again to stop. Your speech is transcribed, run through the active mode, and pasted into the focused app. Escape to cancel.
Every mode can be connected to an AI provider for automatic post-processing.
Or create your own with a custom system prompt.
Raw transcription with no processing. Your words appear exactly as spoken — perfect for drafting emails, notes, or messages.
Automatically fixes grammar, punctuation, and capitalization. Ideal for when you want polished text without editing.
Speak in any language and get translated text. Uses your configured LLM to produce natural, fluent translations.
Dictate a long thought and get back a concise summary. Great for meeting notes, brainstorming, or capturing ideas quickly.
Write your own system prompt, pick an AI provider and model, assign a shortcut, and choose an emoji. Your mode works exactly like the built-in ones.
Three transcription backends, each with different trade-offs.
Switch between them instantly in settings.
The macOS native speech recognition engine. Already on your Mac, no downloads required.
Best for: Quick notes & simple dictation
OpenAI's state-of-the-art model via faster-whisper. Choose from Tiny to Large v3 depending on your accuracy/speed needs.
Best for: Multilingual & high-accuracy work
NVIDIA's 1.1B parameter Parakeet model running natively through ONNX Runtime. No Python, no external dependencies.
Best for: Fast English transcription
Configure record, stop, and cancel hotkeys. Assign per-mode shortcuts with any modifier combo.
Pick different system sounds for start, stop, and transcription complete. Or turn them off entirely.
Connect OpenAI, Anthropic, Claude CLI, or any compatible API. Use different models per mode.
Auto-paste toggle, overlay display, appearance theme, and Python venv path for Whisper.
No. All transcription engines run entirely on your Mac. Your audio never leaves your machine. If you use an AI mode with a cloud provider (like OpenAI or Anthropic), only the transcribed text is sent for post-processing — never the audio itself.
Apple Speech is the easiest — zero setup, works offline. Whisper offers the best accuracy and supports 10+ languages. Parakeet ONNX is the fastest for English and runs natively without Python.
Not for basic transcription. API keys are only needed if you want AI post-processing modes (Clean up, Translate, etc.). You can connect OpenAI, Anthropic, or any OpenAI-compatible local server like Ollama.
Yes. Spkn works system-wide with a global keyboard shortcut. After transcription, the text is pasted into whatever app is focused — text editors, browsers, Slack, email, terminal, anything that accepts ⌘V.
Spkn requires macOS 14 Sonoma or later. It runs natively on both Apple Silicon and Intel Macs.
Absolutely. Create custom modes with your own system prompt, choose an AI provider and model, assign a dedicated keyboard shortcut, and pick an emoji icon. Your custom modes work exactly like the built-in ones.
Yes. Spkn is built with Swift and is fully open source. You can inspect the code, contribute, or modify it to fit your needs.
Spkn is a one-time purchase of $5. No subscription, no hidden fees. You get all features and future updates included.
One-time purchase. No subscription. Runs entirely on your Mac.
$5one-time
Buy Spkn — $5Requires macOS 14 Sonoma or later · Universal binary (Apple Silicon + Intel)