Open Source · Privacy First

Open Source · Local
AI Voice Input

Speak to type, audio to text in seconds. Powered by FunASR, data never leaves your computer.

Homebrew — coming soon: brew install --cask murmur · Winget — coming soon: winget install TeFuirnever.Murmur

11+ AI Models

100% Local

5 Export Formats

Apache-2.0 OSS

Why Murmur

Why Murmur?

Fully Local

Built-in FunASR model. All audio processed on your device. Zero data upload, zero privacy concerns.

Chinese Accuracy

Paraformer-large optimized for Chinese. Outperforms Whisper and macOS Dictation in Chinese scenarios.

AI Polish

Remove filler words, fix mistakes, reformat text. Supports 11+ AI models including Ollama for local processing.

Global Hotkey

Press Cmd+Shift+Space to start, speak, press again. Text appears at your cursor automatically.

Features

Core Features

High-Accuracy Recognition

FunASR Paraformer-large with FSMN-VAD and CT-Transformer for production-grade Chinese ASR.

AI Smart Polish

Auto-remove filler words, correct mistakes, and clean up spoken text with your preferred AI model.

Global Hotkey

Cmd+Shift+Space to start/stop recording. Works system-wide, auto-pastes text at cursor position.

Privacy First

All processing happens locally. Your voice data never leaves your computer. No cloud dependency.

11+ AI Models

OpenAI, DeepSeek, Qwen, GLM, Ollama, LM Studio and more. Pick your provider, Murmur fills in the rest.

File Transcription

Import wav/mp3/m4a/flac files. Export to TXT, SRT, VTT, Markdown, or Word documents.

Transcription History

SQLite storage with full-text search. Browse, search, and export your past transcriptions anytime.

Bilingual Support

Full Chinese and English interface. Optimized for Chinese speech recognition.

Getting Started

Get Started in 3 Steps

01

Install

Download for your platform, or use Homebrew / Winget. Double-click to install.

02

Speak

Press the global hotkey and start talking. Murmur captures your voice with FunASR.

03

AI Polish (Optional)

Configure an AI model in Settings. Your spoken words become clean, polished text automatically.

Comparison

How Does Murmur Compare?

Murmur macOS DictationiFlytekWhisper Desktop
Chinese Accuracy ⭐⭐⭐⭐⭐ ⭐⭐⭐ ⭐⭐⭐⭐⭐ ⭐⭐⭐
Fully Local
AI Post-processing
Open Source & Free
11+ AI Models
Custom Prompts

AI Providers

Supported AI Models

Choose your provider, Murmur handles the rest

OpenAI Cloud
DeepSeek Cloud
Qwen Cloud
GLM Cloud
Groq Cloud
Moonshot Cloud
SiliconFlow Cloud
MiniMax Cloud
OpenRouter Cloud
Ollama Local
LM Studio Local

FAQ

Frequently Asked Questions

macOS says 'cannot verify developer' and won't open the app? +

Right-click the Murmur app → select 'Open' → click 'Open' again. This is macOS's security prompt for unsigned apps. Murmur is free open-source software without a paid developer certificate.

First launch is slow — what's being downloaded? +

On first launch, Murmur downloads the FunASR speech recognition model (~1GB). Once downloaded, it's cached locally and won't need downloading again.

How do I configure AI text optimization? +

Open Settings → enter your API Key in the AI section → select a model provider (DeepSeek, Qwen, Ollama, etc.) → save. AI optimization is optional — speech recognition works without it.

Where is my data stored? +

All data is stored locally in SQLite: macOS ~/Library/Application Support/murmur/transcriptions.db, Windows %APPDATA%/murmur/transcriptions.db. Nothing is uploaded to any server.

What export formats are supported? +

Export to TXT, SRT (subtitles), VTT (Web subtitles), Markdown, and DOCX (Word document).

What is the global hotkey? +

Default: Cmd+Shift+Space (macOS) / Ctrl+Shift+Space (Windows/Linux). Press to start recording, press again to stop and auto-paste text at cursor.

Do I need ffmpeg installed? +

Usually no. Since v1.0.0, Murmur uses Python librosa/soundfile for audio conversion. ffmpeg is only needed as a fallback for mp3/m4a files when librosa is unavailable. WAV/FLAC always work without ffmpeg.

Which platforms are supported? +

macOS (Apple Silicon arm64) and Windows. Linux builds are not yet available. On macOS, the app is unsigned — right-click → Open on first launch.

Ready to Start Voice Input?

Download Murmur, configure in minutes, speak to type today.