100% local — no data leaves your machine

Dictation that just works

Your voice, your machine, your data. murmur runs entirely on-device — no cloud, no accounts, no audio ever leaves your computer.

Get Started
$ curl -sSf https://raw.githubusercontent.com/jafreck/murmur/main/scripts/install.sh | bash
1. Hold Ctrl+Shift+Space
2. 🎙️ "Schedule a meeting with the design team for Thursday at 2pm"
3. Release the hotkey
Schedule a meeting with the design team for Thursday at 2pm

Why murmur?

Fast, private, and works everywhere you type.

🔒

Completely Private

Audio is transcribed on-device and discarded immediately. Zero network requests. Your voice data never leaves your machine.

Fast Transcription

Optimized inference with Metal, CUDA, and Vulkan GPU acceleration. INT4/INT8 quantized models for faster speed with minimal accuracy loss.

🖥️

Cross-Platform

Native support for macOS, Windows, and Linux (X11 & Wayland). Runs as a system tray daemon that starts at login.

🎯

Works Everywhere

Text appears wherever your cursor is — browsers, editors, terminals, chat apps. Uses clipboard paste for universal compatibility.

🎛️

Configurable

Choose your hotkey, model size, language, and dictation mode. Simple JSON config with sensible defaults.

🌍

Multilingual

Multiple ASR engines to choose from — up to 90+ languages with Whisper, highest accuracy with Qwen3-ASR, or fast formatted English with Parakeet.

🧠

Context Aware

murmur detects the active application and adjusts formatting — Markdown in editors, plain text in chat, and more.

Choose Your Engine

murmur supports multiple ASR backends — all running locally on your machine. Models are downloaded from HuggingFace on first run.

Qwen3-ASR Recommended

ModelSizeWERStreamingBest For
1.7b (INT4)~4 GB4.20%✅ NativeLowest error rate

Whisper

ModelSizeSpeedBest For
tiny.en75 MBFastestQuick notes
base.en142 MBFastLightweight English
large-v3-turbo1.6 GBModerate90+ languages

Parakeet-TDT

ModelSizeSpeedBest For
0.6b-v2 (INT8)~500 MBFastEnglish with auto-punctuation

Runs on Your Platform

Native support with GPU acceleration where available.

🍎

macOS

Metal GPU acceleration on Apple Silicon. Globe key as default hotkey.

🪟

Windows

CUDA & Vulkan GPU support. Installs as a background service.

🐧

Linux

X11 and Wayland. CUDA & Vulkan GPU acceleration.

Install in Seconds

Pre-built binaries, no build tools required.

# One-line install — downloads the right binary for your platform curl -sSf https://raw.githubusercontent.com/jafreck/murmur/main/scripts/install.sh | bash # Then start dictating murmur start
# PowerShell — run as administrator irm https://raw.githubusercontent.com/jafreck/murmur/main/scripts/install.ps1 | iex # Then start dictating murmur start
# Clone and build git clone https://github.com/jafreck/murmur.git cd murmur cargo build --release --features metal # Whisper only cargo build --release --features metal,onnx # + Qwen3-ASR & Parakeet # GPU acceleration cargo build --release --features metal # macOS Apple Silicon cargo build --release --features cuda # NVIDIA cargo build --release --features vulkan # Cross-vendor