Local-first voice-to-text companion

Speak anywhere.
Text everywhere.

Hold a hotkey, speak naturally, and AI-polished text appears wherever your cursor is - all processed locally on your machine. No cloud. No account. No friction.

Available for macOS, Windows, and Linux

Loqui app dashboard showing transcription activity chart, stats (550 words, 25 transcriptions, ~14min saved), and recent transcripts from Slack, Notion, Mail and VSCodium

Everything you need to dictate

Loqui handles the entire flow - from voice capture to polished text - without leaving your keyboard.

Hold-to-Dictate

System-wide hotkey works in any app. Press, speak, release - text appears at your cursor instantly.

100% Local Processing

Whisper.cpp with Metal GPU acceleration. Your audio never leaves your device - ever.

Smart Text Cleanup

Optional LLM refactoring cleans up filler words and structures your text. Use local Qwen3 or bring your own key.

Context-Aware Profiles

Loqui detects which app you're in and adapts the writing style - casual for Slack, formal for email.

Always-On Widget

A tiny floating indicator lives on all desktops - even fullscreen. Audio-reactive bars show recording state.

Full History & Stats

Every transcription saved and searchable. Track your dictation habits, time saved, and costs.

Three steps. Zero friction.

From thought to text in seconds - Loqui stays out of your way.

1

Hold

Press your hotkey anywhere on your system. Default is Alt+Space - fully customizable.

2

Speak

Talk naturally in any of 99+ languages. Whisper processes your speech locally in real time.

3

Done

Polished text appears exactly where your cursor is. No copy-paste, no switching apps.

Privacy-first

Your voice never leaves your device

Loqui runs Whisper.cpp entirely on your machine with Metal GPU acceleration. Speech recognition happens locally - no servers, no uploads, no data collection.

Need AI text cleanup? Use the local Qwen3 model or bring your own API key for OpenAI, Anthropic, or Google. Your keys, your choice.

No cloud
No account
No telemetry

Loved by real* people

Don't just take our word for it - take theirs.
*results may vary on the "real" part

Frequently asked questions

Everything you need to know about Loqui.

What speech models does Loqui support?
Loqui uses Whisper.cpp with four model tiers: Tiny (fastest, ~1GB), Base, Small, and Medium (most accurate, ~5GB). All models run locally with Metal GPU acceleration on macOS. You choose the right trade-off between speed and accuracy during setup.
Does it work offline?
Yes - speech-to-text works 100% offline. The Whisper model runs entirely on your machine. The optional "Smart Cleanup" feature (which polishes your transcribed text with an LLM) can also work offline using the local Qwen3 model. Remote LLMs (OpenAI, Anthropic, Google) require internet, but they're strictly opt-in and bring-your-own-key.
Which apps does it work with?
Loqui works system-wide - any app where you can type, you can dictate. Slack, VS Code, Mail, Notion, your browser, terminal, text editors - everywhere. The text is injected at your cursor position via simulated keyboard input.
Is it really free?
Loqui is free and open source. There's no subscription, no freemium, no tracking. If you want to use remote LLMs for text cleanup, you'll need your own API key, but the core dictation feature is completely free.
Which platforms are supported?
Loqui is available for macOS (Apple Silicon), Windows (x64), and Linux (x64 - AppImage, .deb, and .rpm). macOS gets the best performance thanks to Metal GPU acceleration for Whisper.
How is this different from macOS Dictation?
Unlike built-in dictation, Loqui processes everything locally (no cloud), supports context-aware writing profiles that adapt to your current app, includes optional AI text cleanup, and keeps a full searchable history of every transcription.
Why the name "Loqui"?
Loqui (pronounced LOH-kwee) comes from the Latin verb meaning "to speak" - the root behind words like eloquent, loquacious, colloquy, and soliloquy. It felt fitting for an app whose entire purpose is turning speech into text. The Romans had a word for the purest form of speaking - not shouting, not whispering, just speaking. That's exactly what Loqui does: you speak naturally, and it handles the rest.

Ready to speak?

Download Loqui and start dictating in seconds. Free, private, and open source.

Download Loqui