Offline Dictation App
Your voice never touches a server. lowercase processes speech entirely on your device using the NVIDIA Parakeet model — no cloud, no internet, no exceptions.
Why offline dictation matters
Complete privacy
Your voice recordings and transcriptions stay on your hardware. No audio is uploaded, no transcripts are stored on remote servers, and no third party ever processes your speech.
Zero latency
Cloud dictation adds 200–800ms of round-trip network delay. Offline processing on Apple Neural Engine completes in under a second with no network dependency.
Works everywhere
On a plane, in a subway tunnel, at a cabin in the mountains — dictation works identically whether you have gigabit WiFi or no signal at all.
Always reliable
No server outages, no rate limits, no API deprecations. The model lives on your device and works as long as your device does.
How offline processing works
Most dictation apps send your audio to a server, wait for the server to process it, and send text back. This requires an active internet connection and means your voice data passes through third-party infrastructure.
lowercase takes a fundamentally different approach. The NVIDIA Parakeet TDT model — a ~500 MB neural network — is downloaded once and stored locally on your device. When you speak, audio is captured by your microphone, processed by the model running on Apple's Neural Engine (on Mac and iOS) or the device's speech engine (on Android), and the resulting text is pasted into your active app. The entire pipeline happens in local memory.
Because the model processes your complete utterance at once rather than word-by-word, it understands context. It knows whether you meant "their," "there," or "they're" based on the full sentence. This is the same architecture that powers cloud dictation services — but running privately on your own hardware.
Press F5
Activate dictation with a single hotkey
Speak
Talk naturally — no special commands needed
On-device AI
Parakeet model transcribes locally
Text appears
Auto-pasted into your active app
Cloud dictation vs. offline dictation
Every major dictation service — Google Voice Typing, Apple's Siri-based dictation, Microsoft's online speech services — relies on cloud servers to process your audio. Here's what that means in practice, and how lowercase compares.
Offline on every platform
macOS
NVIDIA Parakeet TDT runs on Apple Neural Engine. Press F5 to dictate into any app — even with WiFi off. Requires macOS 14+ and Apple Silicon.
iOS
The same Parakeet model runs on your iPhone's Neural Engine via a custom keyboard extension. Dictate into Messages, Notes, email — any text field, fully offline.
Android
Uses Android's on-device SpeechRecognizer with a floating bubble overlay. Tap, speak, and text is injected into your current app via Accessibility Service.
Windows (Web)
Chrome and Edge provide speech recognition via the Web Speech API. Visit lowercase.click, press F5, and start dictating — no install or account needed.
When offline dictation saves you
Offline isn't just a technical feature — it changes where and how you can use dictation. Think about the situations where cloud services fail you:
Travel
Airplane mode, spotty hotel WiFi, international roaming — dictate emails and notes without worrying about connectivity.
Remote work
Coffee shops with bad WiFi, co-working spaces with captive portals, rural locations — your dictation never depends on the network.
Sensitive environments
Law offices, medical practices, and government buildings where data cannot leave the premises. Offline dictation meets the strictest compliance requirements.
Learn more
Dictate anywhere — no internet needed
Download lowercase and start dictating offline on Mac, iPhone, Android, or in your browser.